- Email: [email protected]
Factors Associated With Potentially Preventable Hospitalization Among Nursing Home Residents in New York State With Chronic Kidney Disease
JAMDA 13 (2012) 337e343
JAMDA journal homepage: www.jamda.com
Original Study
Factors Associated With Potentially Preventable Hospitalization Among Nursing Home Residents in New York State With Chronic Kidney Disease Roy Mathew MD a, b, *, Yuchi Young DrPH c, Srishti Shrestha MPH c a
Stratton VA Medical Center, Albany, NY Albany Medical College, Albany, NY c State University of New York (SUNY) at Albany, School of Public Health, Rensselaer, NY b
a b s t r a c t Keywords: Ambulatory care sensitive hospitalization nursing homes resident determinants chronic kidney disease
Objective: Identify clinical and organizational factors associated with potentially preventable ambulatory care sensitive (ACS) hospitalization among nursing home residents with chronic kidney disease. Methods: New York State Nursing home residents (n ¼ 5449) age 60þ with chronic kidney disease and were hospitalized in 2007. Data included residents’ sociodemographic and clinical characteristics, nursing home organizational factors, and ACS hospitalizations. Multivariate logistic regression quantified the association between potential determinants and ACS hospitalizations (yes versus no). Results: Prevalence of chronic kidney disease among nursing home residents is 24%. Potentially avoidable ACS hospitalization among older nursing home residents with chronic kidney disease is 27%. Three potentially modifiable factors associated with significantly higher odds of ACS hospitalization include the following: presence of congestive heart failure (OR ¼ 1.4; 95% CI 1.24e1.65), excessive medication use (OR ¼ 1.3; 95% CI 1.11e1.48), and the lack of training provided to nursing staff on how to communicate effectively with physician about the resident’s condition. (OR ¼ 1.3; 95% CI 0.59e0.96). Conclusion: To reduce potentially preventable ACS hospitalization among chronic kidney disease patients, congestive heart failure and excessive medication use can be kept stable using relatively simple interventions by periodic multidisciplinary review of medications and assessing appropriate response to therapy; and communication training be provided to nursing staff on how to articulate to the responsible physician important changes in the patients’ condition. Published by Elsevier Inc. on behalf of the American Medical Directors Association, Inc.
Nursing home (NH) residents (NHR) represent a highly vulnerable proportion of the long term care (LTC) population. Interceding illnesses that require acute hospitalization delay the recovery process and result in prolonged or permanent disability.1 Acute hospitalizations not only increase the cost of care but also place the patient at risk for iatrogenic illnesses.2 Previous studies suggested that about 11% to 40% of hospitalizations were potentially avoidable depending on study designs.3,4 Grabowski and colleagues2 examined ambulatory care sensitive hospitalization (ACSH) and cost using New York State NH data from 1999 to 2004 and showed that 29% of all hospitalizations in 2004 met criteria for ACSH; the costs associated with these hospitalizations was $223.8 million. They also noted that a few conditions accounted for most ambulatory case sensitive (ACS) diagnoses and spending for ACSH (eg, pneumonia accounted for 33% of ACSH and 37% of ACSH spending).2
The authors report no conflicts of interest. * Address correspondence to Roy Mathew, MD, Stratton VA Medical Center, 113 Holland Avenue, Albany, NY 12208. E-mail address: [email protected] (R. Mathew).
Intrator et al4 examined the facility and market factors associated with ACSH in NH facilities throughout the United States and found that approximately 32% of all hospitalizations met the criteria for ACSH. To minimize the possible adverse consequence of these hospitalizations and avert unnecessary cost incurred, efforts to identify determinants of potentially avoidable hospitalizations are necessary. Previous studies have attempted to identify important areas for improvement at the facility level, focusing on health manpower, organization structure, and cost associated with ACSH.4e9 Young and associates6,8 examined factors associated with ACSH and found that organizational factors, as well as the patient’s health insurance and family involvement, were more strongly associated with ACSH than individual clinical characteristics. Identifying such factors may allow changes that will affect most patients with the potential for large cost savings. On the other hand, few studies have examined the association between patient conditions and ACSH. Patients with significant medical illness may require specific interventions that are more resource intensive to prevent hospitalizations, above and beyond organizational factors. To target interventions that can
1525-8610/$ - see front matter Published by Elsevier Inc. on behalf of the American Medical Directors Association, Inc. doi:10.1016/j.jamda.2011.01.001
338
R. Mathew et al. / JAMDA 13 (2012) 337e343
further reduce costs associated with preventable hospitalizations it is important to next focus on specific conditions that affect large numbers of patients and require significant resources to maintain stability. Chronic kidney disease (CKD) is one such condition that is highly prevalent in the elderly and commonly associated with ACS conditions in the ambulatory setting.10 Most of the early research on CKD in the elderly was focused on epidemiological characterization of community-dwelling older adults. In an analysis by Stevens and colleagues,10 44% of participants older than 65 years in 2 surveillance cohorts (National Health and Nutrition Examination Survey and Kidney Disease Early Evaluation Program) were identified as having CKD. Go and colleagues11 found that patients with CKD in the community have required higher rates of acute hospital care. In a large cohort from Northern California they found that the need for hospitalization was nearly 3 times higher in those with severe impairment of renal function as compared with those with normal renal function.11 NH residents are frequently excluded from such analyses. The literature regarding CKD in the NH population is limited. McClellan and colleagues12 performed one of the first comprehensive observational analyses of the prevalence of CKD in a national representative NH sample, and found the presence of CKD to be 49%. In addition, patients with CKD were shown to have a high burden of comorbidity including anemia and cardiovascular disease.12 Robinson et al13 found similar results (43.1%) in a large cross-sectional analysis of NHR. These studies have demonstrated that CKD is a highly prevalent condition in the elderly, including those who are NHR, with high disease burden, and require significant health care resource utilization. Our study aims to identify the sociodemographic, clinical, and organizational factors associated with potentially preventable ACSH among the older NHR with CKD. Methods The study protocol was approved by the Health Research Science Board of the New York State Department of Health and the Institutional Review Board at the State University of New York (SUNY) at Albany, New York. Data Source and Study Sample Three data sets were used for this study: New York State Nursing Home Survey, Minimum Data Set (MDS), and Statewide Planning and Research Cooperative System (SPARCS). Data collection methods of 3 data sets have been previously published.6e8 Study Sample This study grew out of larger work undertaken as part of the Nursing Home Study.6 The study population consisted of 5449 residents residing in 146 nursing homes in New York State, aged 60 years and older, with chronic kidney failure, and hospitalized between January 1, 2007, and December 31, 2007. Information on kidney failure was obtained through a nursing home resident assessment and care screening assessment form in the MDS, and ICD-codes 585.x and 586 in the SPARCS hospitalization data. Using both data, the original sample identified meeting our selection criteria was 5903. Of those, 454 were younger than 60 and were excluded from the study, as they possibly possessed different sociodemographic and health characteristics than the older sample. Thus, our final sample for the study consisted of 5449 residents. In the multivariate logistic regression, owing to missing values of various predictor variables, only 4119 residents were included in the final model. Homogeneity tests were performed to compare
those with complete data (n ¼ 4119) and missing data (n ¼ 1330); the results indicated there were no significant differences in age (P ¼ .07) and gender (P ¼ .3). However, there was higher proportion of African Americans with missing predictor values (21% versus 19%) and the difference is statistically significant (P ¼ .04). Outcome Variable A total of 24 ACS diagnoses were used as a proxy variable for potentially preventable hospitalizations; these 24 conditions, if diagnosed early, can be treated in an ambulatory setting, hence avoiding hospitalization.9 ACS diagnoses considered here included angina, asthma, cellulitis, chronic obstructive pulmonary disease, congestive heart failure, dehydration, diabetes mellitus, gastroenteritis, epilepsy, hypertension, hypoglycemia, urinary tract infection, pneumonia, and severe ear, nose, and throat infection.2 These ACS diagnoses were developed for community-dwelling elderly, and have also been applied to and validated in the nursing home population.14 The ACS hospitalization was a dichotomous variable with 1 ¼ potentially preventable ACS hospitalization (24 ACS diagnoses) and 0 ¼ not potentially preventable hospitalization (any diagnosis other than the 24 ACS diagnoses). Information on ACS diagnoses was obtained from the SPARCS data. Predictor Variables Predictor variables used in this study were in 3 groups: sociodemographic and clinical characteristics of the residents and organizational factors of the nursing home. These data were obtained from the Nursing Home Survey and MDS data. Sociodemographic characteristics include age, gender, marital status, race, and education attainment. Age was grouped into 0 ¼ 60e74, 1¼ 75e84, and 2 ¼ 85þ; gender (0 ¼ male, 1 ¼ female); marital status (0 ¼ never married, 1 ¼ widowed, 2 ¼ divorced, 3 ¼ married, 4 ¼ separated); race (0 ¼ White, 1 ¼ African American, 2 ¼ Others); and education attainments (0 ¼ less than high school, 1 ¼ high school graduate, 2 ¼ some college, 3 ¼ Bachelor’s degree/ graduate school). Clinically related factors included were grouped into (1) chronic conditions, (2) use of medication, (3) functional limitation, and (4) social activities. Chronic medical conditions included congestive heart failure, dementia including Alzheimer’s disease, coronary artery disease, and hypertension. Chronic medical conditions were dichotomized with 1 indicating the presence of the disease and 0 otherwise. Number of medications used was a dichotomous variable (0 ¼ 12 or 1 ¼ >12). Functional limitations included 6 activities of daily living (ADL) deficits (dress, eat, transfer, toilet, bath, and walk); a summary variable was created with scores ranging from 0 to 6 (0 ¼ independent of ADL function, 6 ¼ totally dependent). Cognitive skills for daily decision making were 0 ¼ independent, 1 ¼ modified independence only, and 2 ¼ moderately or 3 ¼ severely impaired. In the final analysis, moderately and severely impaired were combined owing to the small number in the severely impaired group. Organizational factors included 5 factors that have been previously studied6e8 or are conceptually relevant to ACSH for CKD: 1. Whether or not training was provided to nursing staff on how to communicate effectively with physicians about a resident’s condition (0 ¼ No versus 1 ¼ Yes). 2. Access to Stat laboratory, radiology, and EKG results in less than 4 hours on weekends (0 ¼ No access, 1 ¼ Difficult but possible, 2 ¼ Easy access). 3. Proportion of residents enrolled in managed care plans/health maintenance organizations (HMOs) or managed long-term
R. Mathew et al. / JAMDA 13 (2012) 337e343
care program for their regular medical care (in percent). The questionnaire asked “What percentage of your residents are enrolled in managed care plans or HMOs for their regular medical care?” The response options were 1 ¼ 0% to 10%, 2 ¼ 11% to 25%, 3 ¼ 26% to 50%, 4 ¼ 51% to 75%, 5 ¼ greater than 75%. 4. How important is the likelihood that illness will cause death when deciding to hospitalize a resident (a 5-point scale; 1 ¼ most unimportant, 5 ¼ most important). 5. Physicians had better access to prior medical history (1 ¼ strongly disagree, 5 ¼ strongly agree).
339
staff on how to effectively communicate with physicians about their residents’ conditions (OR ¼ 1.3; 95% CI 0.59e0.96). Four factors associated with a decreased likelihood of ACSH included African American race as compared with white race (OR ¼ 0.69, 95% CI 0.57e0.85); the presence of hypertension (OR ¼ 0.81, 95% CI 0.69e0.95); the presence of moderate to severe impairment in cognitive skills for daily decision making compared with the reference group (OR ¼ 0.80, 95% CI 0.68e0.96); NH with 75%þ of residents with managed care insurance (OR ¼ 0.49, 95% CI 0.32e0.74). Age was not significantly associated with ACSH. Discussion
Data Analysis Frequency distributions and summary statistics described the dataset. Univariate analyses were performed to describe the nursing home characteristics using percentages and means. Bivariate analysis was performed to examine the difference between ACS and non-ACS hospitalization on selected predictor variables. The predictors of ACS hospitalization were identified using simple logistic regression. Variables associated with preventable hospitalizations that were significant at less than the 0.20 level, clinically relevant, or published in literature were retained for subsequent multivariate logistic regression analysis. Multivariate logistic regression model quantified the association between sociodemographic, clinically related and nursing home organizational factors, and potentially preventable ACS hospitalizations. All statistical tests were conducted using SAS version 9.1 (SAS Institute Inc, Cary, NC) software. Results The prevalence of CKD in our cohort was 24% (Figure 1). Of the 5449 NHR with CKD, 27% (n ¼ 1459) were hospitalized for an ACS diagnosis. Table 1 presents bivariate results of sociodemographic, clinical, and facility characteristics of the study sample. Patients with CKD with an ACSH were significantly older than those without ACSH (81.8 years versus 80.3 years; P < .0001). The ACSH group had a comparatively greater proportion of patients with age older than 85 years than the non-ACSH group (42.7% versus 36.5%). There were more females (60.5% versus 55%; P ¼ .0003) and fewer African Americans (15.3% versus 21%; P < .0001) in the group with ACSH than in the group without ACSH. There were no significant differences in education attainment or facility characteristics between the 2 groups. The most prevalent chronic medical conditions, of the conditions available, were hypertension (40.5%) and congestive heart failure (CHF) (76.4%). Patients with CHF had comparatively higher ACSH (48.9% versus 37.3%), whereas hypertensive patients were less likely to have ACSH (74.2% versus 77.2%) compared with the reference group. There is an equally high level of functional disability in both groups as indicated by mean number of activities of daily living (ADL) deficits (4.21 versus 4.24; P ¼ .43). For prescribed medication use, patients with ACSH were prescribed more medications than those without ACSH (13 versus 12, P < .0001). Multivariate logistic regression revealed 8 factors that were significantly associated with ACSH in patients with CKD (Table 2). Four factors that were associated with an increased likelihood of ACSH included (all values presented as odds ratio [OR], 95% confidence interval [CI]) female gender (OR ¼ 1.28, 95% CI 1.11e1.48); the presence of CHF as a comorbidity (OR ¼ 1.43, 95% CI 1.24e1.65); greater number of medications prescribed (>12) at the NH (OR ¼ 1.28, 95% CI 1.11e1.48); and lack of training provided to the nursing
We decided to focus our analysis on the CKD population because of the established high prevalence in the elderly and the need for greater understanding of the effects on the more vulnerable subsets of this group: the LTC population. Furthermore, CKD frequently coexists with other significant medical problems (eg, diabetes and CHF), thus increasing the complexity of its management. In an effort to expand our knowledge base of the effects of CKD in the LTC population, our study results contribute to the understanding of CKD and associated ACS hospitalization in NHR in 4 ways: (1) provide information on prevalence of CKD in NHR requiring acute hospital transfer and of the proportion of these hospitalizations that may be preventable (ACSH), (2) identify factors associated with ACSH, (3) identify modifiable factors associated with ACSH, and (4) propose intervention strategies to reduce ACS hospitalization among CKD residents. Our findings of CKD prevalence in NHR of 24% is lower than that reported by 2 previous studies. McClellan and colleagues identified 49% of the NH population to have an estimated glomerular filtration rate (eGFR) lower than 60, thus meeting the criteria for CKD.12 Robinson and colleagues13 examined the available laboratory and clinical data of a large database of NHR in the United States. They found that 43.1% of the entire cohort met the eGFR definition for CKD. They also noted a significant association with age and the prevalence of CKD: 33.1% of those 65 to 74 as opposed to 49% in those older than 85.13 However, the difference in prevalence may be because our cohort included only those NHR that required acute hospital transfer whereas the previous studies included all NHR. After analysis for potential areas for ACSH reduction specific to patients with CKD, we identified 3 potentially modifiable factors that correlated with ACSH. Two of the clinical factors were CHF and greater number of medications prescribed. It is expected that patients with CHF would more often have an ACSH, as CHF exacerbation is one of the ACS diagnoses. CKD patients with CHF as comorbidity have a 43% higher risk of ACSH than CKD patients without CHF conditions (Table 1). However, the relationship between CKD and CHF bears special attention. The cardio-renal syndrome has been used to describe a complex state where dysfunction in either organ (heart or kidney) leads to further dysfunction in the other.15 The presence of CKD has been shown to increase the risk of hospitalization and mortality in patients with CHF.16 Patients with the dual diagnoses of CHF and CKD will require more frequent monitoring so that acute hospitalizations are minimized. Following implementation of a guideline-based, multidisciplinary care model for patients admitted to an acute-care facility with acute decompensated CHF, 30-day readmission was reduced from 6.4% to 6.0% (not significant), and in-hospital mortality was reduced from 5.8% to 2.1% (P < .05).17 In the NH setting, this may be accomplished with physicians or midlevel providers conducting routine dedicated rounds for patients with these high-risk conditions. Two important factors identified by Ouslander and colleagues18 as “very or somewhat helpful” in preventing avoidable hospitalizations are availability of physician or physician extender
340
R. Mathew et al. / JAMDA 13 (2012) 337e343
Fig. 1. Sample selection. NYS, New York State; SPARCS, statewide planning and research cooperative system; MDS, minimum data set; NHS, nursing home survey; DON, director of nursing; CKD, chronic kidney disease; ICD-9, International Statistical Classification of Diseases and Related Health Problemse9th Edition; ACS, ambulatory care sensitive diagnosis; ACSH, ambulatory care sensitive hospitalization.
at least 3-days per week and regular availability of nurse practitioner. On the other hand, a systematic review of the literature to date by Grabowski et al3 suggests that the association between provider availability and hospitalization is not consistent across studies. However, the authors state that the context of potential benefit (eg, higher risk patients) needs to be analyzed in greater detail. The number of medications prescribed to patients with CKD is another modifiable risk factor that is associated with higher odds of ACSH in our analysis. Multiple medications are likely to be prescribed for the multitude of comorbid conditions present in patients with CKD (eg, angiotensin receptor blockers for the prevention of progression of diabetic nephropathy of type II diabetes mellitus, aspirin and HMG-coA reductase inhibitors [“Statins”] for cardiovascular risk reduction). In a recent review by Rifkin and Winkelmayer,19 it was noted that elderly individuals in the United States are taking an average of 5 prescription and nonprescription medications. The average number of medications listed on the medication sheets for our study sample is 12. The potential for drug-drug interactions at this level of medication burden increases exponentially, especially if additional symptomatic medications (eg, pain relievers, sleep aids) are prescribed. These medications require frequent monitoring and special attention is needed for medications cleared by the kidney. The aid of a dedicated pharmacist with specialization in geriatric pharmacology may provide added assistance to the provider and minimize the possibility of significant adverse events (ie, nonsteroidal anti-inflammatory drugs and angiotensin-converting enzyme inhibitors inducing acute kidney injury).20,21 Diuretic use is a special case in point of a medication requiring careful monitoring in patients with CKD, despite not retaining significance in our analysis (data not shown). The use of high doses of diuretics (Furosemide) in the setting of CKD was associated with worsening renal function during index hospitalization for heart failure, as well as with subsequent heart
failure hospitalization or cardiovascular-related death, in a prospective analysis of heart failure admissions.16 Simple use of daily weights and early specialist referral if diuretic efficacy is in question may help prevent worsening renal function and the need for acute hospital transfer. Of the organizational factors, we found that NH that did not provide training to nursing staff on how to communicate effectively with physicians on patients’ conditions had higher odds of ACSH in patients with CKD. This finding is consistent with previous studies.6,8,22 Young and associates6 found, among other predictors, that training provided to nursing staff on how to communicate effectively with physicians about a resident’s condition is significantly associated with lower rate of ACS hospitalizations per 100,000 resident-days than those who did not receive training, and that translates to 3% fewer potentially preventable hospitalizations and $6,714,000 in cost savings. We would suggest that an organizational intervention to provide communication training to nursing staff on how to gauge and articulate a patient’s condition be implemented in NHs. Timely management of problems in high-risk individuals requires prompt and complete notification of the responsible provider by front-line personnel, which is often nurses in NHs. Another interesting finding of this study is that moderately or severely impaired cognitive skills for daily decision making is associated with lower odds of ACSH than the reference group. This finding is contradictory to a previous study in which Ouslander and colleagues found NHs with high hospitalization rates had a higher percentage of patients with severe cognitive impairment.18 However, the medical reviewers who conducted the data review and surveys for the Ouslander et al analysis felt that in 20% to 29% of preventable hospitalizations, it was “important or somewhat important” that the providers should have considered that the resident’s overall condition indicated that the resident would not benefit from an acute hospital transfer. In support of this and our
R. Mathew et al. / JAMDA 13 (2012) 337e343
341
Table 1 Comparison Between Ambulatory Care Sensitive (ACS) and Non-ACS Hospitalizations for Sociodemographics, Clinical, and Nonclinical Characteristics of the Study Population Among Older Nursing Home Residents in New York State Characteristics of Residents Sociodemographics Age 60e74 years 75e84 years >85 years Mean age (SD) in years Range Gender Female Male Marital status Never married Widowed Divorced Married Separated Race/Ethnicity Caucasian African American Others Education Below high school High school Some college/Technical education Bachelor’s degree and above Chronic medical conditions Congestive heart failure No Yes Dementia including Alzheimer’s disease No Yes Coronary artery disease No Yes Hypertension No Yes Number and type of medications Mean (SD) Range Summary ADL deficits (SD) (0 to 6) Cognitive skills for daily decision making Independent Modified independence only Moderately impaired Severely impaired Nonclinical conditions Nursing staff in this facility provided with training on how to communicate effectively with physicians about the resident’s condition No Yes Stat laboratory results in <4 hours on weekends Easy access Difficult but possible No access Percentage of the residents are enrolled in MCPs or HMOs for their regular medical care 0% to 10% 11% to 25% 26% to 50% 51% to 75% 52% to 75% Likelihood that illness will cause death Most unimportant Unimportant Neutral Important Most important Physicians had better access to prior medical history 1 ¼ Strongly disagree 2 ¼ Disagree
Total Sample (N ¼ 5449)
No ACS (n ¼ 3990)
ACS (n ¼ 1459)
*P Value
1487 (27.29%) 1882 (34.54%) 2080 (38.17%) 80.72 (9.42) 60 to 110
1137 (28.5%) 1396 (34.99%) 1457 (36.52%) 80.30 (9.45) 60e110
350 (23.99%) 486 (33.31%) 623 (42.7%) 81.85 (9.24) 60e102
<.0001
3076 (56.45%) 2373 (43.55%)
2194 (54.99%) 1796 (45.01%)
882 (60.45%) 577 (39.55%)
.0003
844 1445 2359 121 350
642 1078 1657 97 253
202 367 702 24 97
(14.51%) (26.36%) (50.43%) (1.72%) (6.97%)
.0012
1063 (74.39%) 218 (15.26%) 148 (10.36%)
<.0001
(16.49%) (28.23%) (46.08%) (2.36%) (6.84%)
(17.23%) (28.92%) (44.46%) (2.6%) (6.79%)
3699 (68.92%) 1046 (19.49%) 622 (11.59%)
2636 (66.94%) 828 (21.03%) 474 (12.04%)
1109 1509 417 324
801 1115 289 247
(33.02%) (44.92%) (12.41%) (9.65%)
(32.67%) (45.47%) (11.79%) (10.07%)
308 394 128 77
(33.96%) (43.44%) (14.11%) (8.49%)
<.0001
.14
3076 (59.52%) 2092 (40.48%)
2361 (62.66%) 1407 (37.34%)
715 (51.07%) 685 (48.93%)
<.0001
3331 (68.84%) 1508 (31.16%)
2399 (67.88%) 1135 (32.12%)
932 (71.42%) 373 (28.58%)
.0185
4173 (80.75%) 995 (19.25%)
3081 (81.77%) 687 (18.23%)
1142 (23.6%) 3697 (76.4%)
805 (22.78%) 2729 (77.22%)
12.45 (4.57) 0 to 36 4.23 (1.08)
12.29 (4.54) 0 to 36 4.24 (1.07)
1951 1457 1537 354
1404 1026 1163 279
(36.82%) (27.5%) (29.01%) (6.68%)
(36.26%) (26.5%) (30.04%) (7.21%)
1092 (78%) 308 (22%) 337 (25.82%) 968 (74.18%) 12.92 (4.65) 1 to 31 4.21 (1.12) 547 431 374 75
(38.33%) (30.2%) (26.21%) (5.26%)
.0023
.03
<.0001 .43 .001
4324 (89.32%) 517 (10.68%)
3146 (88.92%) 392 (11.08%)
1178 (90.41%) 125 (9.59%)
.14
2092 (43.5%) 2435 (50.63%) 282 (5.86%)
1525 (43.47%) 1798 (51.25%) 185 (5.27%)
567 (43.58%) 637 (48.96%) 97 (7.46%)
.28
2872 1275 327 196 190
(59.09%) (26.23%) (6.73%) (4.03%) (3.91%)
2080 962 225 135 158
(58.43%) (27.02%) (6.32%) (3.79%) (4.44%)
792 313 102 61 32
(60.92%) (24.08%) (7.85%) (4.69%) (2.46%)
.13
30 410 520 1280 2559
(0.63%) (8.54%) (10.84%) (26.67%) (53.32%)
19 328 342 921 1923
(0.54%) (9.28%) (9.68%) (26.07%) (54.43%)
11 82 178 359 636
(0.87%) (6.48%) (14.06%) (28.36%) (50.24%)
.22
355 (28.06%) 412 (32.57%)
.90
1444 (30.13%) 1327 (27.69%)
1089 (30.87%) 915 (25.94%)
(continued on next page)
342
R. Mathew et al. / JAMDA 13 (2012) 337e343
Table 1 (continued ) Characteristics of Residents 3 ¼ Uncertain 4 ¼ Agree 5 ¼ Strongly agree
Total Sample (N ¼ 5449) 849 (17.71%) 787 (16.42%) 386 (8.05%)
No ACS (n ¼ 3990) 662 (18.76%) 591 (16.75%) 271 (7.68%)
ACS (n ¼ 1459)
*P Value
187 (14.78%) 196 (15.49%) 115 (9.09%)
All values presented as n(%). *Chi squared test or Student t test P value (ACS versus Non ACS).
findings, the systematic review by Grabowski et al3 revealed that a consistent association between cognitive impairment and lower hospitalization rates is seen in a number of prior studies. Why moderately and severely impaired cognitive skills for daily decision making has 20% lower odds of ACSH compared with those with independence in decision making is not clear. This may suggest that providers are considering the potential for long-term benefit of acute transfer. In other words, patients with little chance for full cognitive/functional recovery would benefit less from aggressive therapy for acute illnesses. The summation of our findings points to the need for comprehensive disease management as a means to prevent escalation of problems to the point of requiring acute hospitalization. This may be best provided by a multidisciplinary team involving, at the least, a primary provider, a nurse, and a pharmacist. Comprehensive care for patients with complicated or numerous chronic conditions has been found to reduce excessive medical resource utilization and in some cases to reduce medical expenditure. A comprehensive review of the various models was conducted by Boult and colleagues.23 For example, interdisciplinary team management and
individualized disease management has been shown to improve quality of life and reduce hospital utilization for CHF and COPD, respectively.23 However, it has been recognized that the widespread use of these techniques may be limited because of the increased costs incurred when hiring the needed personnel, as well as acquiring the additional resources required.24 Therefore, limiting such resource-intensive disease management strategies to subsets of patients who would derive the greatest benefit could potentially minimize the expenses involved. Our current findings also supplement our previous analyses using this dataset. Some of the factors have been identified as significant factors in the prior analyses: for example, higher percentage of NHR with managed care insurance as having a lower odds of ACSH is consistent with previous studies.6,8 However, the prior analyses focused on the LTC population as a whole (in demographics and underlying medical conditions), and organizational factors were the most important predictors of ACSH. We explored the possibility that identifying a commonly present chronic condition would highlight specific clinical areas of modification to prevent ACSH; and indeed clinical conditions were
Table 2 Predictors of Ambulatory Care Sensitive Hospitalization among Older Nursing Home Residents in New York State: Multivariable Logistic Regression (N ¼ 4119)
Sociodemographics Age 60e74 years 75e84 years >85 years Gender Male Female Race/Ethnicity Caucasian African American Others Clinical conditions Congestive heart failure No Yes Hypertension No Yes Cognitive skills for daily decision making Independent Modified independence only Moderately or severely impaired Number of medications 12 >12 Nonclinical conditions Provide training on how to communicate effectively with physicians about the resident’s condition Yes No Percentage of the residents enrolled in MCPs or HMOs for their regular medical care 0 to 10% 11 to 25% 26 to 50% 51 to 75% 75% above MCP, managed care programs; HMO, health maintenance organization.
Unadjusted Odds Ratio (95% CI)
Adjusted Odds Ratio (95% CI)
Reference 1.13 (0.97e1.33) 1.39 (1.19e1.62)
Reference 0.98 (0.81e1.18) 1.13 (0.93e1.38)
Reference 1.25 (1.11e1.41)
Reference 1.28 (1.11e1.48)
Reference 0.65 (0.55e0.77) 0.77 (0.64e0.94)
Reference 0.69 (0.57e0.85) 0.83 (0.65e1.05)
Reference 1.61 (1.42e1.82)
Reference 1.43 (1.24e1.65)
Reference 0.85 (0.73e0.98)
Reference 0.81 (0.69e0.95)
Reference 1.08 (0.93e1.25) 0.80 (0.69e0.92)
Reference 0.99 (0.83e1.19) 0.80 (0.68e0.96)
Reference 1.30 (1.15e1.47)
Reference 1.28 (1.11e1.48)
Reference 0.85 (0.69e1.05)
Reference 1.33 (0.59e0.96)
Reference 0.85 (0.73e0.99) 1.19 (0.93e1.53) 1.19 (0.87e1.62) 0.53 (0.36e0.79)
Reference 0.87 (0.74e1.03) 1.10 (0.83e1.47) 1.11 (0.72e1.45) 0.49 (0.32e0.74)
R. Mathew et al. / JAMDA 13 (2012) 337e343
prominent predictors of ACSH, in contrast to our previous analyses. As mentioned, it would be difficult, and likely unnecessary, to specifically address these clinical problems for all individuals. Potential Limitations A number of limitations need to be considered in the interpretation of our analysis. The 3 datasets used for our analysis contain comprehensive information to better understand sociodemograpics, health status, functional limitations, and hospitalizations of NHR as well as NH facility characteristics. However, these datasets are not specifically collected for our study and do not necessarily contain all relevant information needed to address our research questions. Only a few medication classes were available, and pertinent laboratory variables were not available for analysis. An additional factor that requires attention is the proper identification of elderly individuals with CKD. There were 2968 of 5449 patients with CKD in our cohort identified in SPARCS with a diagnostic code for CKD, but not noted to have “kidney failure” in the MDS. We also were unable to accurately distinguish between CKD patients on hemodialysis and those who are predialysis, as well as any possible kidney transplant recipients. We suggest that specific definitions of CKD be included in the tracking mechanism (MDS) for NHR. Despite these limitations, our study bridged the gaps in the literature by providing additional data about the prevalence of CKD in NHR and factors related to ACSH in patients with CKD. Conclusions Patients with CKD make up a considerable proportion of those NHR requiring transition to a higher level of care. We have identified several areas for intervention to minimize the need for acute hospital transfer. To effectively target the modifiable factors identified in this study (CHF, medication burden, nurse-physician communication), we suggest a multidisciplinary team (including nursing staff, provider, and a pharmacist) approach to the medical management of NHR to minimize ACSH. Prospective intervention using a multidisciplinary team to prevent unnecessary hospitalizations among NHR with CKD is recommended. References 1. Creditor MC. Hazards of hospitalization of the elderly. Ann Intern Med 1993; 118:219e223. 2. Grabowski DC, O’Malley AJ, Barhydt NR. The costs and potential savings associated with nursing home hospitalizations. Health Aff (Millwood) 2007;26: 1753e1761.
343
3. Grabowski DC, Stewart KA, Broderick SA, Coots LA. Predictors of nursing home hospitalization: A review of the literature. Med Car Res Rev 2008;65:3e39. 4. Intrator O, Zinn J, Mor V. Nursing home characteristics and potentially preventable hospitalizations of long-stay residents. J Am Geriatr Soc 2004;52: 1730e1736. 5. Carter MW. Factors associated with ambulatory care: Sensitive hospitalizations among nursing home residents. J Aging Health 2003;15:295e331. 6. Young Y, Barhydt NR, Broderick S, et al. Factors associated with potentially preventable hospitalization in nursing home residents in New York State: A survey of directors of nursing. J Am Geriatr Soc 2010;58:901e907. 7. Young Y, Inamdar S, Barhydt NR, et al. Preventable hospitalization among nursing home residents: Varying views between medical directors and directors of nursing regarding determinants. J Aging Health 2010;22:169e182. 8. Young Y, Inamdar S, Dichter BS, et al. Clinical and nonclinical factors associated with potentially preventable hospitalizations among nursing home residents in New York State. J Am Med Dir Assoc; 2010. doi:10.1016/j.jamda.2010.03.006. 9. Culler SD, Parchman ML, Przybylski M. Factors related to potentially preventable hospitalizations among the elderly. Med Care 1998;36:804e817. 10. Stevens LA, Li S, Wang C, et al. Prevalence of CKD and comorbid illness in elderly patients in the United States: Results from the Kidney Early Evaluation Program (KEEP). Am J Kidney Dis 2010;55:S23e33. 11. Go AS, Chertow GM, Fan D, et al. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med 2004;351:1296e1305. 12. McClellan WM, Resnick B, Bradbury BD, et al. Prevalence and severity of chronic kidney disease and anemia in the nursing home population. J Am Med Dir Assoc 2010;11:33e41. 13. Robinson B, Artz AS, Culleton B, et al. Prevalence of anemia in the nursing home: Contribution of chronic kidney disease. J Am Geriatr Soc 2007;55: 1566e1570. 14. Carter MW, Porell FW. Variations in hospitalization rates among nursing home residents: The role of facility and market attributes. Gerontologist 2003;43: 175e191. 15. Ronco C, Haapio M, House AA, et al. Cardiorenal syndrome. J Am Coll Cardiol 2008;52:1527e1539. 16. Metra M, Nodari S, Parrinello G, et al. Worsening renal function in patients hospitalised for acute heart failure: Clinical implications and prognostic significance. Eur J Heart Fail 2008;10:188e195. 17. Costantini O, Huck K, Carlson MD, et al. Impact of a guideline-based disease management team on outcomes of hospitalized patients with congestive heart failure. Arch Intern Med 2001;161:177e182. 18. Ouslander JG, Lamb G, Perloe M, et al. Potentially avoidable hospitalizations of nursing home residents: Frequency, causes, and costs. J Am Geriatr Soc 2010; 58:627e635. 19. Rifkin DE, Winkelmayer WC. Medication issues in older individuals with CKD. Adv Chronic Kidney Dis 2010;17:320e328. 20. Halvorsen KH, Ruths S, Granas AG, Viktil KK. Multidisciplinary intervention to identify and resolve drug-related problems in Norwegian nursing homes. Scand J Prim Health Care 2010;28:82e88. 21. Verrue CL, Petrovic M, Mehuys E, et al. Pharmacists’ interventions for optimization of medication use in nursing homes: A systematic review. Drugs Aging 2009;26:37e49. 22. Tjia J, Mazor KM, Field T, et al. Nurse-physician communication in the longterm care setting: Perceived barriers and impact on patient safety. J Patient Saf 2009;5:145e152. 23. Boult C, Green AF, Boult LB, et al. Successful models of comprehensive care for older adults with chronic conditions: Evidence for the Institute of Medicine’s “retooling for an aging America” report. J Am Geriatr Soc 2009; 57:2328e2337. 24. Bradley EH, Webster TR, Baker D, et al. After adoption: Sustaining the innovation. A case study of disseminating the hospital elder life program. J Am Geriatr Soc 2005;53:1455e1461.
JAMDA journal homepage: www.jamda.com
Original Study
Factors Associated With Potentially Preventable Hospitalization Among Nursing Home Residents in New York State With Chronic Kidney Disease Roy Mathew MD a, b, *, Yuchi Young DrPH c, Srishti Shrestha MPH c a
Stratton VA Medical Center, Albany, NY Albany Medical College, Albany, NY c State University of New York (SUNY) at Albany, School of Public Health, Rensselaer, NY b
a b s t r a c t Keywords: Ambulatory care sensitive hospitalization nursing homes resident determinants chronic kidney disease
Objective: Identify clinical and organizational factors associated with potentially preventable ambulatory care sensitive (ACS) hospitalization among nursing home residents with chronic kidney disease. Methods: New York State Nursing home residents (n ¼ 5449) age 60þ with chronic kidney disease and were hospitalized in 2007. Data included residents’ sociodemographic and clinical characteristics, nursing home organizational factors, and ACS hospitalizations. Multivariate logistic regression quantified the association between potential determinants and ACS hospitalizations (yes versus no). Results: Prevalence of chronic kidney disease among nursing home residents is 24%. Potentially avoidable ACS hospitalization among older nursing home residents with chronic kidney disease is 27%. Three potentially modifiable factors associated with significantly higher odds of ACS hospitalization include the following: presence of congestive heart failure (OR ¼ 1.4; 95% CI 1.24e1.65), excessive medication use (OR ¼ 1.3; 95% CI 1.11e1.48), and the lack of training provided to nursing staff on how to communicate effectively with physician about the resident’s condition. (OR ¼ 1.3; 95% CI 0.59e0.96). Conclusion: To reduce potentially preventable ACS hospitalization among chronic kidney disease patients, congestive heart failure and excessive medication use can be kept stable using relatively simple interventions by periodic multidisciplinary review of medications and assessing appropriate response to therapy; and communication training be provided to nursing staff on how to articulate to the responsible physician important changes in the patients’ condition. Published by Elsevier Inc. on behalf of the American Medical Directors Association, Inc.
Nursing home (NH) residents (NHR) represent a highly vulnerable proportion of the long term care (LTC) population. Interceding illnesses that require acute hospitalization delay the recovery process and result in prolonged or permanent disability.1 Acute hospitalizations not only increase the cost of care but also place the patient at risk for iatrogenic illnesses.2 Previous studies suggested that about 11% to 40% of hospitalizations were potentially avoidable depending on study designs.3,4 Grabowski and colleagues2 examined ambulatory care sensitive hospitalization (ACSH) and cost using New York State NH data from 1999 to 2004 and showed that 29% of all hospitalizations in 2004 met criteria for ACSH; the costs associated with these hospitalizations was $223.8 million. They also noted that a few conditions accounted for most ambulatory case sensitive (ACS) diagnoses and spending for ACSH (eg, pneumonia accounted for 33% of ACSH and 37% of ACSH spending).2
The authors report no conflicts of interest. * Address correspondence to Roy Mathew, MD, Stratton VA Medical Center, 113 Holland Avenue, Albany, NY 12208. E-mail address: [email protected] (R. Mathew).
Intrator et al4 examined the facility and market factors associated with ACSH in NH facilities throughout the United States and found that approximately 32% of all hospitalizations met the criteria for ACSH. To minimize the possible adverse consequence of these hospitalizations and avert unnecessary cost incurred, efforts to identify determinants of potentially avoidable hospitalizations are necessary. Previous studies have attempted to identify important areas for improvement at the facility level, focusing on health manpower, organization structure, and cost associated with ACSH.4e9 Young and associates6,8 examined factors associated with ACSH and found that organizational factors, as well as the patient’s health insurance and family involvement, were more strongly associated with ACSH than individual clinical characteristics. Identifying such factors may allow changes that will affect most patients with the potential for large cost savings. On the other hand, few studies have examined the association between patient conditions and ACSH. Patients with significant medical illness may require specific interventions that are more resource intensive to prevent hospitalizations, above and beyond organizational factors. To target interventions that can
1525-8610/$ - see front matter Published by Elsevier Inc. on behalf of the American Medical Directors Association, Inc. doi:10.1016/j.jamda.2011.01.001
338
R. Mathew et al. / JAMDA 13 (2012) 337e343
further reduce costs associated with preventable hospitalizations it is important to next focus on specific conditions that affect large numbers of patients and require significant resources to maintain stability. Chronic kidney disease (CKD) is one such condition that is highly prevalent in the elderly and commonly associated with ACS conditions in the ambulatory setting.10 Most of the early research on CKD in the elderly was focused on epidemiological characterization of community-dwelling older adults. In an analysis by Stevens and colleagues,10 44% of participants older than 65 years in 2 surveillance cohorts (National Health and Nutrition Examination Survey and Kidney Disease Early Evaluation Program) were identified as having CKD. Go and colleagues11 found that patients with CKD in the community have required higher rates of acute hospital care. In a large cohort from Northern California they found that the need for hospitalization was nearly 3 times higher in those with severe impairment of renal function as compared with those with normal renal function.11 NH residents are frequently excluded from such analyses. The literature regarding CKD in the NH population is limited. McClellan and colleagues12 performed one of the first comprehensive observational analyses of the prevalence of CKD in a national representative NH sample, and found the presence of CKD to be 49%. In addition, patients with CKD were shown to have a high burden of comorbidity including anemia and cardiovascular disease.12 Robinson et al13 found similar results (43.1%) in a large cross-sectional analysis of NHR. These studies have demonstrated that CKD is a highly prevalent condition in the elderly, including those who are NHR, with high disease burden, and require significant health care resource utilization. Our study aims to identify the sociodemographic, clinical, and organizational factors associated with potentially preventable ACSH among the older NHR with CKD. Methods The study protocol was approved by the Health Research Science Board of the New York State Department of Health and the Institutional Review Board at the State University of New York (SUNY) at Albany, New York. Data Source and Study Sample Three data sets were used for this study: New York State Nursing Home Survey, Minimum Data Set (MDS), and Statewide Planning and Research Cooperative System (SPARCS). Data collection methods of 3 data sets have been previously published.6e8 Study Sample This study grew out of larger work undertaken as part of the Nursing Home Study.6 The study population consisted of 5449 residents residing in 146 nursing homes in New York State, aged 60 years and older, with chronic kidney failure, and hospitalized between January 1, 2007, and December 31, 2007. Information on kidney failure was obtained through a nursing home resident assessment and care screening assessment form in the MDS, and ICD-codes 585.x and 586 in the SPARCS hospitalization data. Using both data, the original sample identified meeting our selection criteria was 5903. Of those, 454 were younger than 60 and were excluded from the study, as they possibly possessed different sociodemographic and health characteristics than the older sample. Thus, our final sample for the study consisted of 5449 residents. In the multivariate logistic regression, owing to missing values of various predictor variables, only 4119 residents were included in the final model. Homogeneity tests were performed to compare
those with complete data (n ¼ 4119) and missing data (n ¼ 1330); the results indicated there were no significant differences in age (P ¼ .07) and gender (P ¼ .3). However, there was higher proportion of African Americans with missing predictor values (21% versus 19%) and the difference is statistically significant (P ¼ .04). Outcome Variable A total of 24 ACS diagnoses were used as a proxy variable for potentially preventable hospitalizations; these 24 conditions, if diagnosed early, can be treated in an ambulatory setting, hence avoiding hospitalization.9 ACS diagnoses considered here included angina, asthma, cellulitis, chronic obstructive pulmonary disease, congestive heart failure, dehydration, diabetes mellitus, gastroenteritis, epilepsy, hypertension, hypoglycemia, urinary tract infection, pneumonia, and severe ear, nose, and throat infection.2 These ACS diagnoses were developed for community-dwelling elderly, and have also been applied to and validated in the nursing home population.14 The ACS hospitalization was a dichotomous variable with 1 ¼ potentially preventable ACS hospitalization (24 ACS diagnoses) and 0 ¼ not potentially preventable hospitalization (any diagnosis other than the 24 ACS diagnoses). Information on ACS diagnoses was obtained from the SPARCS data. Predictor Variables Predictor variables used in this study were in 3 groups: sociodemographic and clinical characteristics of the residents and organizational factors of the nursing home. These data were obtained from the Nursing Home Survey and MDS data. Sociodemographic characteristics include age, gender, marital status, race, and education attainment. Age was grouped into 0 ¼ 60e74, 1¼ 75e84, and 2 ¼ 85þ; gender (0 ¼ male, 1 ¼ female); marital status (0 ¼ never married, 1 ¼ widowed, 2 ¼ divorced, 3 ¼ married, 4 ¼ separated); race (0 ¼ White, 1 ¼ African American, 2 ¼ Others); and education attainments (0 ¼ less than high school, 1 ¼ high school graduate, 2 ¼ some college, 3 ¼ Bachelor’s degree/ graduate school). Clinically related factors included were grouped into (1) chronic conditions, (2) use of medication, (3) functional limitation, and (4) social activities. Chronic medical conditions included congestive heart failure, dementia including Alzheimer’s disease, coronary artery disease, and hypertension. Chronic medical conditions were dichotomized with 1 indicating the presence of the disease and 0 otherwise. Number of medications used was a dichotomous variable (0 ¼ 12 or 1 ¼ >12). Functional limitations included 6 activities of daily living (ADL) deficits (dress, eat, transfer, toilet, bath, and walk); a summary variable was created with scores ranging from 0 to 6 (0 ¼ independent of ADL function, 6 ¼ totally dependent). Cognitive skills for daily decision making were 0 ¼ independent, 1 ¼ modified independence only, and 2 ¼ moderately or 3 ¼ severely impaired. In the final analysis, moderately and severely impaired were combined owing to the small number in the severely impaired group. Organizational factors included 5 factors that have been previously studied6e8 or are conceptually relevant to ACSH for CKD: 1. Whether or not training was provided to nursing staff on how to communicate effectively with physicians about a resident’s condition (0 ¼ No versus 1 ¼ Yes). 2. Access to Stat laboratory, radiology, and EKG results in less than 4 hours on weekends (0 ¼ No access, 1 ¼ Difficult but possible, 2 ¼ Easy access). 3. Proportion of residents enrolled in managed care plans/health maintenance organizations (HMOs) or managed long-term
R. Mathew et al. / JAMDA 13 (2012) 337e343
care program for their regular medical care (in percent). The questionnaire asked “What percentage of your residents are enrolled in managed care plans or HMOs for their regular medical care?” The response options were 1 ¼ 0% to 10%, 2 ¼ 11% to 25%, 3 ¼ 26% to 50%, 4 ¼ 51% to 75%, 5 ¼ greater than 75%. 4. How important is the likelihood that illness will cause death when deciding to hospitalize a resident (a 5-point scale; 1 ¼ most unimportant, 5 ¼ most important). 5. Physicians had better access to prior medical history (1 ¼ strongly disagree, 5 ¼ strongly agree).
339
staff on how to effectively communicate with physicians about their residents’ conditions (OR ¼ 1.3; 95% CI 0.59e0.96). Four factors associated with a decreased likelihood of ACSH included African American race as compared with white race (OR ¼ 0.69, 95% CI 0.57e0.85); the presence of hypertension (OR ¼ 0.81, 95% CI 0.69e0.95); the presence of moderate to severe impairment in cognitive skills for daily decision making compared with the reference group (OR ¼ 0.80, 95% CI 0.68e0.96); NH with 75%þ of residents with managed care insurance (OR ¼ 0.49, 95% CI 0.32e0.74). Age was not significantly associated with ACSH. Discussion
Data Analysis Frequency distributions and summary statistics described the dataset. Univariate analyses were performed to describe the nursing home characteristics using percentages and means. Bivariate analysis was performed to examine the difference between ACS and non-ACS hospitalization on selected predictor variables. The predictors of ACS hospitalization were identified using simple logistic regression. Variables associated with preventable hospitalizations that were significant at less than the 0.20 level, clinically relevant, or published in literature were retained for subsequent multivariate logistic regression analysis. Multivariate logistic regression model quantified the association between sociodemographic, clinically related and nursing home organizational factors, and potentially preventable ACS hospitalizations. All statistical tests were conducted using SAS version 9.1 (SAS Institute Inc, Cary, NC) software. Results The prevalence of CKD in our cohort was 24% (Figure 1). Of the 5449 NHR with CKD, 27% (n ¼ 1459) were hospitalized for an ACS diagnosis. Table 1 presents bivariate results of sociodemographic, clinical, and facility characteristics of the study sample. Patients with CKD with an ACSH were significantly older than those without ACSH (81.8 years versus 80.3 years; P < .0001). The ACSH group had a comparatively greater proportion of patients with age older than 85 years than the non-ACSH group (42.7% versus 36.5%). There were more females (60.5% versus 55%; P ¼ .0003) and fewer African Americans (15.3% versus 21%; P < .0001) in the group with ACSH than in the group without ACSH. There were no significant differences in education attainment or facility characteristics between the 2 groups. The most prevalent chronic medical conditions, of the conditions available, were hypertension (40.5%) and congestive heart failure (CHF) (76.4%). Patients with CHF had comparatively higher ACSH (48.9% versus 37.3%), whereas hypertensive patients were less likely to have ACSH (74.2% versus 77.2%) compared with the reference group. There is an equally high level of functional disability in both groups as indicated by mean number of activities of daily living (ADL) deficits (4.21 versus 4.24; P ¼ .43). For prescribed medication use, patients with ACSH were prescribed more medications than those without ACSH (13 versus 12, P < .0001). Multivariate logistic regression revealed 8 factors that were significantly associated with ACSH in patients with CKD (Table 2). Four factors that were associated with an increased likelihood of ACSH included (all values presented as odds ratio [OR], 95% confidence interval [CI]) female gender (OR ¼ 1.28, 95% CI 1.11e1.48); the presence of CHF as a comorbidity (OR ¼ 1.43, 95% CI 1.24e1.65); greater number of medications prescribed (>12) at the NH (OR ¼ 1.28, 95% CI 1.11e1.48); and lack of training provided to the nursing
We decided to focus our analysis on the CKD population because of the established high prevalence in the elderly and the need for greater understanding of the effects on the more vulnerable subsets of this group: the LTC population. Furthermore, CKD frequently coexists with other significant medical problems (eg, diabetes and CHF), thus increasing the complexity of its management. In an effort to expand our knowledge base of the effects of CKD in the LTC population, our study results contribute to the understanding of CKD and associated ACS hospitalization in NHR in 4 ways: (1) provide information on prevalence of CKD in NHR requiring acute hospital transfer and of the proportion of these hospitalizations that may be preventable (ACSH), (2) identify factors associated with ACSH, (3) identify modifiable factors associated with ACSH, and (4) propose intervention strategies to reduce ACS hospitalization among CKD residents. Our findings of CKD prevalence in NHR of 24% is lower than that reported by 2 previous studies. McClellan and colleagues identified 49% of the NH population to have an estimated glomerular filtration rate (eGFR) lower than 60, thus meeting the criteria for CKD.12 Robinson and colleagues13 examined the available laboratory and clinical data of a large database of NHR in the United States. They found that 43.1% of the entire cohort met the eGFR definition for CKD. They also noted a significant association with age and the prevalence of CKD: 33.1% of those 65 to 74 as opposed to 49% in those older than 85.13 However, the difference in prevalence may be because our cohort included only those NHR that required acute hospital transfer whereas the previous studies included all NHR. After analysis for potential areas for ACSH reduction specific to patients with CKD, we identified 3 potentially modifiable factors that correlated with ACSH. Two of the clinical factors were CHF and greater number of medications prescribed. It is expected that patients with CHF would more often have an ACSH, as CHF exacerbation is one of the ACS diagnoses. CKD patients with CHF as comorbidity have a 43% higher risk of ACSH than CKD patients without CHF conditions (Table 1). However, the relationship between CKD and CHF bears special attention. The cardio-renal syndrome has been used to describe a complex state where dysfunction in either organ (heart or kidney) leads to further dysfunction in the other.15 The presence of CKD has been shown to increase the risk of hospitalization and mortality in patients with CHF.16 Patients with the dual diagnoses of CHF and CKD will require more frequent monitoring so that acute hospitalizations are minimized. Following implementation of a guideline-based, multidisciplinary care model for patients admitted to an acute-care facility with acute decompensated CHF, 30-day readmission was reduced from 6.4% to 6.0% (not significant), and in-hospital mortality was reduced from 5.8% to 2.1% (P < .05).17 In the NH setting, this may be accomplished with physicians or midlevel providers conducting routine dedicated rounds for patients with these high-risk conditions. Two important factors identified by Ouslander and colleagues18 as “very or somewhat helpful” in preventing avoidable hospitalizations are availability of physician or physician extender
340
R. Mathew et al. / JAMDA 13 (2012) 337e343
Fig. 1. Sample selection. NYS, New York State; SPARCS, statewide planning and research cooperative system; MDS, minimum data set; NHS, nursing home survey; DON, director of nursing; CKD, chronic kidney disease; ICD-9, International Statistical Classification of Diseases and Related Health Problemse9th Edition; ACS, ambulatory care sensitive diagnosis; ACSH, ambulatory care sensitive hospitalization.
at least 3-days per week and regular availability of nurse practitioner. On the other hand, a systematic review of the literature to date by Grabowski et al3 suggests that the association between provider availability and hospitalization is not consistent across studies. However, the authors state that the context of potential benefit (eg, higher risk patients) needs to be analyzed in greater detail. The number of medications prescribed to patients with CKD is another modifiable risk factor that is associated with higher odds of ACSH in our analysis. Multiple medications are likely to be prescribed for the multitude of comorbid conditions present in patients with CKD (eg, angiotensin receptor blockers for the prevention of progression of diabetic nephropathy of type II diabetes mellitus, aspirin and HMG-coA reductase inhibitors [“Statins”] for cardiovascular risk reduction). In a recent review by Rifkin and Winkelmayer,19 it was noted that elderly individuals in the United States are taking an average of 5 prescription and nonprescription medications. The average number of medications listed on the medication sheets for our study sample is 12. The potential for drug-drug interactions at this level of medication burden increases exponentially, especially if additional symptomatic medications (eg, pain relievers, sleep aids) are prescribed. These medications require frequent monitoring and special attention is needed for medications cleared by the kidney. The aid of a dedicated pharmacist with specialization in geriatric pharmacology may provide added assistance to the provider and minimize the possibility of significant adverse events (ie, nonsteroidal anti-inflammatory drugs and angiotensin-converting enzyme inhibitors inducing acute kidney injury).20,21 Diuretic use is a special case in point of a medication requiring careful monitoring in patients with CKD, despite not retaining significance in our analysis (data not shown). The use of high doses of diuretics (Furosemide) in the setting of CKD was associated with worsening renal function during index hospitalization for heart failure, as well as with subsequent heart
failure hospitalization or cardiovascular-related death, in a prospective analysis of heart failure admissions.16 Simple use of daily weights and early specialist referral if diuretic efficacy is in question may help prevent worsening renal function and the need for acute hospital transfer. Of the organizational factors, we found that NH that did not provide training to nursing staff on how to communicate effectively with physicians on patients’ conditions had higher odds of ACSH in patients with CKD. This finding is consistent with previous studies.6,8,22 Young and associates6 found, among other predictors, that training provided to nursing staff on how to communicate effectively with physicians about a resident’s condition is significantly associated with lower rate of ACS hospitalizations per 100,000 resident-days than those who did not receive training, and that translates to 3% fewer potentially preventable hospitalizations and $6,714,000 in cost savings. We would suggest that an organizational intervention to provide communication training to nursing staff on how to gauge and articulate a patient’s condition be implemented in NHs. Timely management of problems in high-risk individuals requires prompt and complete notification of the responsible provider by front-line personnel, which is often nurses in NHs. Another interesting finding of this study is that moderately or severely impaired cognitive skills for daily decision making is associated with lower odds of ACSH than the reference group. This finding is contradictory to a previous study in which Ouslander and colleagues found NHs with high hospitalization rates had a higher percentage of patients with severe cognitive impairment.18 However, the medical reviewers who conducted the data review and surveys for the Ouslander et al analysis felt that in 20% to 29% of preventable hospitalizations, it was “important or somewhat important” that the providers should have considered that the resident’s overall condition indicated that the resident would not benefit from an acute hospital transfer. In support of this and our
R. Mathew et al. / JAMDA 13 (2012) 337e343
341
Table 1 Comparison Between Ambulatory Care Sensitive (ACS) and Non-ACS Hospitalizations for Sociodemographics, Clinical, and Nonclinical Characteristics of the Study Population Among Older Nursing Home Residents in New York State Characteristics of Residents Sociodemographics Age 60e74 years 75e84 years >85 years Mean age (SD) in years Range Gender Female Male Marital status Never married Widowed Divorced Married Separated Race/Ethnicity Caucasian African American Others Education Below high school High school Some college/Technical education Bachelor’s degree and above Chronic medical conditions Congestive heart failure No Yes Dementia including Alzheimer’s disease No Yes Coronary artery disease No Yes Hypertension No Yes Number and type of medications Mean (SD) Range Summary ADL deficits (SD) (0 to 6) Cognitive skills for daily decision making Independent Modified independence only Moderately impaired Severely impaired Nonclinical conditions Nursing staff in this facility provided with training on how to communicate effectively with physicians about the resident’s condition No Yes Stat laboratory results in <4 hours on weekends Easy access Difficult but possible No access Percentage of the residents are enrolled in MCPs or HMOs for their regular medical care 0% to 10% 11% to 25% 26% to 50% 51% to 75% 52% to 75% Likelihood that illness will cause death Most unimportant Unimportant Neutral Important Most important Physicians had better access to prior medical history 1 ¼ Strongly disagree 2 ¼ Disagree
Total Sample (N ¼ 5449)
No ACS (n ¼ 3990)
ACS (n ¼ 1459)
*P Value
1487 (27.29%) 1882 (34.54%) 2080 (38.17%) 80.72 (9.42) 60 to 110
1137 (28.5%) 1396 (34.99%) 1457 (36.52%) 80.30 (9.45) 60e110
350 (23.99%) 486 (33.31%) 623 (42.7%) 81.85 (9.24) 60e102
<.0001
3076 (56.45%) 2373 (43.55%)
2194 (54.99%) 1796 (45.01%)
882 (60.45%) 577 (39.55%)
.0003
844 1445 2359 121 350
642 1078 1657 97 253
202 367 702 24 97
(14.51%) (26.36%) (50.43%) (1.72%) (6.97%)
.0012
1063 (74.39%) 218 (15.26%) 148 (10.36%)
<.0001
(16.49%) (28.23%) (46.08%) (2.36%) (6.84%)
(17.23%) (28.92%) (44.46%) (2.6%) (6.79%)
3699 (68.92%) 1046 (19.49%) 622 (11.59%)
2636 (66.94%) 828 (21.03%) 474 (12.04%)
1109 1509 417 324
801 1115 289 247
(33.02%) (44.92%) (12.41%) (9.65%)
(32.67%) (45.47%) (11.79%) (10.07%)
308 394 128 77
(33.96%) (43.44%) (14.11%) (8.49%)
<.0001
.14
3076 (59.52%) 2092 (40.48%)
2361 (62.66%) 1407 (37.34%)
715 (51.07%) 685 (48.93%)
<.0001
3331 (68.84%) 1508 (31.16%)
2399 (67.88%) 1135 (32.12%)
932 (71.42%) 373 (28.58%)
.0185
4173 (80.75%) 995 (19.25%)
3081 (81.77%) 687 (18.23%)
1142 (23.6%) 3697 (76.4%)
805 (22.78%) 2729 (77.22%)
12.45 (4.57) 0 to 36 4.23 (1.08)
12.29 (4.54) 0 to 36 4.24 (1.07)
1951 1457 1537 354
1404 1026 1163 279
(36.82%) (27.5%) (29.01%) (6.68%)
(36.26%) (26.5%) (30.04%) (7.21%)
1092 (78%) 308 (22%) 337 (25.82%) 968 (74.18%) 12.92 (4.65) 1 to 31 4.21 (1.12) 547 431 374 75
(38.33%) (30.2%) (26.21%) (5.26%)
.0023
.03
<.0001 .43 .001
4324 (89.32%) 517 (10.68%)
3146 (88.92%) 392 (11.08%)
1178 (90.41%) 125 (9.59%)
.14
2092 (43.5%) 2435 (50.63%) 282 (5.86%)
1525 (43.47%) 1798 (51.25%) 185 (5.27%)
567 (43.58%) 637 (48.96%) 97 (7.46%)
.28
2872 1275 327 196 190
(59.09%) (26.23%) (6.73%) (4.03%) (3.91%)
2080 962 225 135 158
(58.43%) (27.02%) (6.32%) (3.79%) (4.44%)
792 313 102 61 32
(60.92%) (24.08%) (7.85%) (4.69%) (2.46%)
.13
30 410 520 1280 2559
(0.63%) (8.54%) (10.84%) (26.67%) (53.32%)
19 328 342 921 1923
(0.54%) (9.28%) (9.68%) (26.07%) (54.43%)
11 82 178 359 636
(0.87%) (6.48%) (14.06%) (28.36%) (50.24%)
.22
355 (28.06%) 412 (32.57%)
.90
1444 (30.13%) 1327 (27.69%)
1089 (30.87%) 915 (25.94%)
(continued on next page)
342
R. Mathew et al. / JAMDA 13 (2012) 337e343
Table 1 (continued ) Characteristics of Residents 3 ¼ Uncertain 4 ¼ Agree 5 ¼ Strongly agree
Total Sample (N ¼ 5449) 849 (17.71%) 787 (16.42%) 386 (8.05%)
No ACS (n ¼ 3990) 662 (18.76%) 591 (16.75%) 271 (7.68%)
ACS (n ¼ 1459)
*P Value
187 (14.78%) 196 (15.49%) 115 (9.09%)
All values presented as n(%). *Chi squared test or Student t test P value (ACS versus Non ACS).
findings, the systematic review by Grabowski et al3 revealed that a consistent association between cognitive impairment and lower hospitalization rates is seen in a number of prior studies. Why moderately and severely impaired cognitive skills for daily decision making has 20% lower odds of ACSH compared with those with independence in decision making is not clear. This may suggest that providers are considering the potential for long-term benefit of acute transfer. In other words, patients with little chance for full cognitive/functional recovery would benefit less from aggressive therapy for acute illnesses. The summation of our findings points to the need for comprehensive disease management as a means to prevent escalation of problems to the point of requiring acute hospitalization. This may be best provided by a multidisciplinary team involving, at the least, a primary provider, a nurse, and a pharmacist. Comprehensive care for patients with complicated or numerous chronic conditions has been found to reduce excessive medical resource utilization and in some cases to reduce medical expenditure. A comprehensive review of the various models was conducted by Boult and colleagues.23 For example, interdisciplinary team management and
individualized disease management has been shown to improve quality of life and reduce hospital utilization for CHF and COPD, respectively.23 However, it has been recognized that the widespread use of these techniques may be limited because of the increased costs incurred when hiring the needed personnel, as well as acquiring the additional resources required.24 Therefore, limiting such resource-intensive disease management strategies to subsets of patients who would derive the greatest benefit could potentially minimize the expenses involved. Our current findings also supplement our previous analyses using this dataset. Some of the factors have been identified as significant factors in the prior analyses: for example, higher percentage of NHR with managed care insurance as having a lower odds of ACSH is consistent with previous studies.6,8 However, the prior analyses focused on the LTC population as a whole (in demographics and underlying medical conditions), and organizational factors were the most important predictors of ACSH. We explored the possibility that identifying a commonly present chronic condition would highlight specific clinical areas of modification to prevent ACSH; and indeed clinical conditions were
Table 2 Predictors of Ambulatory Care Sensitive Hospitalization among Older Nursing Home Residents in New York State: Multivariable Logistic Regression (N ¼ 4119)
Sociodemographics Age 60e74 years 75e84 years >85 years Gender Male Female Race/Ethnicity Caucasian African American Others Clinical conditions Congestive heart failure No Yes Hypertension No Yes Cognitive skills for daily decision making Independent Modified independence only Moderately or severely impaired Number of medications 12 >12 Nonclinical conditions Provide training on how to communicate effectively with physicians about the resident’s condition Yes No Percentage of the residents enrolled in MCPs or HMOs for their regular medical care 0 to 10% 11 to 25% 26 to 50% 51 to 75% 75% above MCP, managed care programs; HMO, health maintenance organization.
Unadjusted Odds Ratio (95% CI)
Adjusted Odds Ratio (95% CI)
Reference 1.13 (0.97e1.33) 1.39 (1.19e1.62)
Reference 0.98 (0.81e1.18) 1.13 (0.93e1.38)
Reference 1.25 (1.11e1.41)
Reference 1.28 (1.11e1.48)
Reference 0.65 (0.55e0.77) 0.77 (0.64e0.94)
Reference 0.69 (0.57e0.85) 0.83 (0.65e1.05)
Reference 1.61 (1.42e1.82)
Reference 1.43 (1.24e1.65)
Reference 0.85 (0.73e0.98)
Reference 0.81 (0.69e0.95)
Reference 1.08 (0.93e1.25) 0.80 (0.69e0.92)
Reference 0.99 (0.83e1.19) 0.80 (0.68e0.96)
Reference 1.30 (1.15e1.47)
Reference 1.28 (1.11e1.48)
Reference 0.85 (0.69e1.05)
Reference 1.33 (0.59e0.96)
Reference 0.85 (0.73e0.99) 1.19 (0.93e1.53) 1.19 (0.87e1.62) 0.53 (0.36e0.79)
Reference 0.87 (0.74e1.03) 1.10 (0.83e1.47) 1.11 (0.72e1.45) 0.49 (0.32e0.74)
R. Mathew et al. / JAMDA 13 (2012) 337e343
prominent predictors of ACSH, in contrast to our previous analyses. As mentioned, it would be difficult, and likely unnecessary, to specifically address these clinical problems for all individuals. Potential Limitations A number of limitations need to be considered in the interpretation of our analysis. The 3 datasets used for our analysis contain comprehensive information to better understand sociodemograpics, health status, functional limitations, and hospitalizations of NHR as well as NH facility characteristics. However, these datasets are not specifically collected for our study and do not necessarily contain all relevant information needed to address our research questions. Only a few medication classes were available, and pertinent laboratory variables were not available for analysis. An additional factor that requires attention is the proper identification of elderly individuals with CKD. There were 2968 of 5449 patients with CKD in our cohort identified in SPARCS with a diagnostic code for CKD, but not noted to have “kidney failure” in the MDS. We also were unable to accurately distinguish between CKD patients on hemodialysis and those who are predialysis, as well as any possible kidney transplant recipients. We suggest that specific definitions of CKD be included in the tracking mechanism (MDS) for NHR. Despite these limitations, our study bridged the gaps in the literature by providing additional data about the prevalence of CKD in NHR and factors related to ACSH in patients with CKD. Conclusions Patients with CKD make up a considerable proportion of those NHR requiring transition to a higher level of care. We have identified several areas for intervention to minimize the need for acute hospital transfer. To effectively target the modifiable factors identified in this study (CHF, medication burden, nurse-physician communication), we suggest a multidisciplinary team (including nursing staff, provider, and a pharmacist) approach to the medical management of NHR to minimize ACSH. Prospective intervention using a multidisciplinary team to prevent unnecessary hospitalizations among NHR with CKD is recommended. References 1. Creditor MC. Hazards of hospitalization of the elderly. Ann Intern Med 1993; 118:219e223. 2. Grabowski DC, O’Malley AJ, Barhydt NR. The costs and potential savings associated with nursing home hospitalizations. Health Aff (Millwood) 2007;26: 1753e1761.
343
3. Grabowski DC, Stewart KA, Broderick SA, Coots LA. Predictors of nursing home hospitalization: A review of the literature. Med Car Res Rev 2008;65:3e39. 4. Intrator O, Zinn J, Mor V. Nursing home characteristics and potentially preventable hospitalizations of long-stay residents. J Am Geriatr Soc 2004;52: 1730e1736. 5. Carter MW. Factors associated with ambulatory care: Sensitive hospitalizations among nursing home residents. J Aging Health 2003;15:295e331. 6. Young Y, Barhydt NR, Broderick S, et al. Factors associated with potentially preventable hospitalization in nursing home residents in New York State: A survey of directors of nursing. J Am Geriatr Soc 2010;58:901e907. 7. Young Y, Inamdar S, Barhydt NR, et al. Preventable hospitalization among nursing home residents: Varying views between medical directors and directors of nursing regarding determinants. J Aging Health 2010;22:169e182. 8. Young Y, Inamdar S, Dichter BS, et al. Clinical and nonclinical factors associated with potentially preventable hospitalizations among nursing home residents in New York State. J Am Med Dir Assoc; 2010. doi:10.1016/j.jamda.2010.03.006. 9. Culler SD, Parchman ML, Przybylski M. Factors related to potentially preventable hospitalizations among the elderly. Med Care 1998;36:804e817. 10. Stevens LA, Li S, Wang C, et al. Prevalence of CKD and comorbid illness in elderly patients in the United States: Results from the Kidney Early Evaluation Program (KEEP). Am J Kidney Dis 2010;55:S23e33. 11. Go AS, Chertow GM, Fan D, et al. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med 2004;351:1296e1305. 12. McClellan WM, Resnick B, Bradbury BD, et al. Prevalence and severity of chronic kidney disease and anemia in the nursing home population. J Am Med Dir Assoc 2010;11:33e41. 13. Robinson B, Artz AS, Culleton B, et al. Prevalence of anemia in the nursing home: Contribution of chronic kidney disease. J Am Geriatr Soc 2007;55: 1566e1570. 14. Carter MW, Porell FW. Variations in hospitalization rates among nursing home residents: The role of facility and market attributes. Gerontologist 2003;43: 175e191. 15. Ronco C, Haapio M, House AA, et al. Cardiorenal syndrome. J Am Coll Cardiol 2008;52:1527e1539. 16. Metra M, Nodari S, Parrinello G, et al. Worsening renal function in patients hospitalised for acute heart failure: Clinical implications and prognostic significance. Eur J Heart Fail 2008;10:188e195. 17. Costantini O, Huck K, Carlson MD, et al. Impact of a guideline-based disease management team on outcomes of hospitalized patients with congestive heart failure. Arch Intern Med 2001;161:177e182. 18. Ouslander JG, Lamb G, Perloe M, et al. Potentially avoidable hospitalizations of nursing home residents: Frequency, causes, and costs. J Am Geriatr Soc 2010; 58:627e635. 19. Rifkin DE, Winkelmayer WC. Medication issues in older individuals with CKD. Adv Chronic Kidney Dis 2010;17:320e328. 20. Halvorsen KH, Ruths S, Granas AG, Viktil KK. Multidisciplinary intervention to identify and resolve drug-related problems in Norwegian nursing homes. Scand J Prim Health Care 2010;28:82e88. 21. Verrue CL, Petrovic M, Mehuys E, et al. Pharmacists’ interventions for optimization of medication use in nursing homes: A systematic review. Drugs Aging 2009;26:37e49. 22. Tjia J, Mazor KM, Field T, et al. Nurse-physician communication in the longterm care setting: Perceived barriers and impact on patient safety. J Patient Saf 2009;5:145e152. 23. Boult C, Green AF, Boult LB, et al. Successful models of comprehensive care for older adults with chronic conditions: Evidence for the Institute of Medicine’s “retooling for an aging America” report. J Am Geriatr Soc 2009; 57:2328e2337. 24. Bradley EH, Webster TR, Baker D, et al. After adoption: Sustaining the innovation. A case study of disseminating the hospital elder life program. J Am Geriatr Soc 2005;53:1455e1461.