Advertisement

Provision of Home-Based Primary Care to Individuals With Intellectual and/or Developmental Disability Is Associated With a Lower Hospitalization Rate Than a Traditional Primary Care Model

Open AccessPublished:June 14, 2022DOI:https://doi.org/10.1016/j.jamda.2022.05.011

      Abstract

      Objectives

      The objective of this study was to determine if providing home-based primary care (HBPC) to individuals with intellectual and/or developmental disabilities (IDD) was associated with a lower hospitalization rate than a control group receiving traditional primary care.

      Design and Intervention

      Individuals with IDD living in supported residential settings in Ohio were offered HBPC. Individuals electing HBPC made up the intervention group. Those who did not opt for HBPC continued to receive traditional primary care services and made up the control group. Hospitalizations were tracked in both groups.

      Setting and Participants

      The 757 study participants had IDD diagnoses and received residential support services throughout the study period.

      Methods

      Annualized hospitalization rate was determined in both groups and was compared using generalized estimating equations while controlling for patients’ age and hospitalization rate in the year prior to the study.

      Results

      The results showed that group membership had a significant effect on the hospitalization rate (Wald χ2 = 20.71, P < .01). Being in the control group was associated with a 2.12-fold increase in annual hospitalization rate for a given patient. The overall population hospitalization rate was 329 hospitalizations per 1000 per year in the HBPC-receiving individuals and 619 hospitalizations per 1000 per year in the control group.

      Conclusions and Implications

      We found that individuals with IDD receiving HBPC were hospitalized at a lower rate than a control group receiving traditional primary care. Expanding access to HBPC may be a worthwhile priority for organizations that support individuals with IDD.

      Keywords

      Home-based primary care (HBPC) has been shown to reduce hospitalizations, rehospitalizations, and skilled nursing facility placements and improve patient and caregiver satisfaction.
      • Desai N.R.
      • Smith K.L.
      • Boal J.
      The positive financial contribution of home-based primary care programs: the case of the Mount Sinai Visiting Doctors.
      • Wajnberg A.
      • Wang K.H.
      • Aniff M.
      • et al.
      Hospitalizations and skilled nursing facility placements before and after the implementation of a home-based primary care program.
      • Hughes S.L.
      • Weaver F.M.
      • Giobbie-Hurder A.
      • et al.
      Effectiveness of team-managed home-based primary care.
      • Rotenberg J.
      • Kinosian B.
      • Boling P.
      • Taler G.
      Independence at Home Learning Collaborative Writing Group. Home-based primary care: beyond extension of the independence at home demonstration.
      To date, most HBPC outcomes studies have included Medicare beneficiaries, but other populations may derive substantial benefit as well. For instance, congregate living residents can benefit from on-site HBPC provided by physicians and nonphysician practitioners who are familiar with the residential care setting.
      • Sloane P.D.
      • Zimmerman S.
      • Perez R.
      • et al.
      Physician perspectives on medical care delivery in assisted living.
      ,
      • Mills W.R.
      • Robertson S.P.
      • Ogard K.L.
      • et al.
      Hospital care of assisted living residents by an assisted living based primary care team.
      Individuals with intellectual and/or developmental disabilities (IDD) are increasingly recognized as high-risk for medical complications and death. Nearly 60% of individuals with IDD have 3 or more chronic conditions in addition to their IDD diagnosis.
      United Healthcare
      Quality improvement for individuals with intellectual and developmental disabilities.
      The US IDD population is aging, and the number of persons aged ≥60 years will double to 1.2 million in the next 25 years.
      United Healthcare
      Quality improvement for individuals with intellectual and developmental disabilities.
      Many of these individuals have difficulty accessing a physician's office for care and experience high rates of emergency room and hospital utilization.
      • Durbin A.
      • Balogh R.
      • Lin E.
      • et al.
      Repeat emergency department visits for individuals with intellectual and developmental disabilities and psychiatric disorders.
      ,
      • Durbin A.
      • Balogh R.
      • Lin E.
      • Wilton A.S.
      • Lunsky Y.
      Emergency department use: common presenting issues and continuity of care for individuals with and without intellectual and developmental disabilities.
      They also experience a higher mortality rate than the general population, often due to potentially preventable conditions.
      • O’Leary L.
      • Cooper S.A.
      • Hughes-McCormack L.
      Early death and causes of death of people with intellectual disabilities: a systematic review.
      ,
      • Shooshtari S.
      • Ouellette-Kuntz H.
      • Balogh R.
      • et al.
      Patterns of mortality among adults with intellectual and developmental disabilities in the Canadian province of Manitoba.
      As set forth by the Olmstead Act, individuals with IDD must have access to high-quality services comparable to those available to people without IDD.
      • Davies S.
      • Schmidt E.
      • Schultz E.
      • et al.
      Home and community-based services quality indicators.
      ,
      • Friedman C.
      Building the Framework for IDD Quality Measures.
      Despite evidence of HBPC's benefit in geriatric and medically complex patients, little is known about the effect of the care model on outcomes in individuals with IDD. Hospitalization rate is a value-based outcome measure that is beginning to be applied to the IDD population.
      • Mills W.R.
      • Law K.
      • Boyd R.
      • et al.
      Hospitalization rate and days spent at home as value-based outcome measures for people with intellectual and developmental disability.
      We hypothesized that providing HBPC to individuals with IDD living in residential settings would be associated with a lower hospitalization rate than a control group receiving traditional primary care.

      Methods

      Our study population included individuals in Ohio who resided in a community living setting for individuals with IDD managed by ResCare Community Living (Res-Care, Inc, Louisville, KY). Individuals with IDD were offered HBPC as an alternative for their primary medical care by an affiliated HBPC practice (Western Reserve Medical Group, Cleveland, OH). Two hundred seven individuals with IDD opted for and received the HBPC care model between April 1, 2020, and February 28, 2022. Written consent was obtained from all HBPC participants. Another 550 individuals with IDD who did not opt for HBPC and were living in the other congregate living settings in the state of Ohio served as a control group. Individuals in the control group continued to receive medical care from their existing, traditional, office-based primary care provider. Individuals in both groups received personal care and nursing support services by ResCare Community Living.
      HBPC consisted of regular visits to the community living site by an HBPC physician or nurse practitioner. Patients were visited approximately every 5 weeks, compared with a traditional primary care visit frequency of approximately every 12 weeks. All HBPC services were billed to third-party payors via the physician fee schedule. HBPC visits focused on the evaluation and management of new symptoms, active management of chronic diseases, implementation of recommended preventive medicine services, and an assessment of the home environment. All HBPC patients or their representatives had 24 × 7 access to a medical practitioner from the practice to assist with in-between visit concerns and triage. Eligible patients were also enrolled in chronic care management using structured assessment tools, supplemented with monthly phone calls to the patient or their representative.
      In the HBPC group, total managed days was calculated as the number of days an individual received HBPC. In the control group, total managed days was calculated as the days an individual served was using residential services minus days that the individual was on leave from the program. In the HBPC group, hospitalizations were tracked by the HBPC practice via ping notifications from a statewide health information exchange, and hospitalizations were verified by a practice nurse who contacted the patient's care team. To track hospitalizations in the control group, each individual's daily whereabouts was entered into a database (HCS Interactant, Wall Township, NJ), recording whether the individual was at home, in the community receiving supportive services from our organization, or in other settings, including being in the hospital. In both groups, hospitalization rate per 1000 was calculated as [Hospitalizations/(Managed Days/1000)] × 365.
      • McCall N.
      • Geonnotti K.
      Utilization measures worksheet. Math Policy Res.

      Statistical Analyses

      Our data were collected and analyzed on the individual level. To make statistical comparisons in hospitalization rates between the control and HBPC groups, rates were calculated per individual per year, adjusting for the number of managed days for that individual [(number of hospitalizations per patient/managed days per patient) × 365]. With count and rate data (eg, number of hospital admissions and hospitalization rates), the assumption of multivariate normality of the dependent variable is unlikely, and sophisticated longitudinal data analytic techniques are preferred.
      • Twisk J.W.
      Applied Longitudinal Data Analysis for Epidemiology: A Practical Guide.
      These techniques fall under the umbrella of generalized linear models (GZLM), which provide a flexible approach to the distribution of the dependent variable (eg, binominal, Poisson, or Gamma) and link function that will transform the relationship between the mean of the dependent variable and the linear combination of the predictor variables so that it is linear. The general idea behind GZLM longitudinal analysis techniques such as mixed linear modeling and generalized estimating equations is that the adjustments for correlations of multiple observations taken from the same subject can be performed very efficiently modeling the relationships between the variables in the analysis simultaneously irrespective of the nature of the covariate or the dependent variable (eg, continuous, dichotomous, or categorial). Therefore, we compared the groups' individual annualized hospitalization rates using generalized estimating equations with group (HBPC or control) and sex entered as fixed factors, whereas participant's age and individual hospitalization rate in the year prior to the start of the study were specified as covariates. We also directly statistically compared the groups on each of these predictors to provide baseline sample characteristics. Continuous variables such as number of managed days, age, and individual hospitalization rate in the year prior to the study were analyzed using independent samples t tests with bootstrapping on 1000 samples to correct for statistical bias associated with groups of unequal size and unequal variances within each group. Effect sizes for these group comparisons are reported as Glass Delta (Δ), which substitutes the control group's SD for the population SD, thus standardizing the denominator variable across experimental contrasts: Δ= (μ1 – μ2)/SDControl.
      • Ferguson C.J.
      An effect size primer: A guide for clinicians and researchers.
      Frequency data (ie, sex, primary diagnosis) were analyzed using χ2 tests with bootstrapping.

      Results

      A total of 757 individuals were studied (207 HBPC-receiving and 550 control). During the study period, there were 64,404 total managed days in the HBPC group and 170,432 managed days in the control group. Fifty-eight hospitalizations occurred in the HBPC group, and 289 hospitalizations occurred in the control group, corresponding to a hospitalization rate of 329 hospitalizations per 1000 per year in the HBPC-receiving individuals and 619 hospitalizations per 1000 per year in the control group. As shown in Table 1, there were no statistically significant differences in the proportion of males and females (χ2 = 0.305, P = .58) or the frequency of a particular diagnostic category (ie, mild, moderate, severe, or profound intellectual disability; χ2 = 1.93, P = .59) between HBPC and control groups. As shown in Table 2, the groups did not differ in age (Mcontrol = 49.12, SD = 18.73, vs MHBPC = 48.99, SD = 18.76; P = .94) or in the number of managed days (Mcontrol = 307.56, SD = 106.56, vs MHBPC = 317.56, SD = 191.34; P = .53). The groups, however, did significantly differ in the individual hospitalization rate in the year prior to the beginning of the study (t = 3.79, P < .01). Specifically, the control group experienced an average of 1.02 (SD = 3.66) hospitalizations per person per year, whereas in the HBPC group the mean hospitalization rate was 0.32 (SD = 1.29). Because of this unexpected baseline difference between groups, we built a statistical model to control for baseline hospitalization rate in order to appropriately measure the effect of the HBPC intervention.
      Table 1Frequencies and χ2 Test Comparisons Between the Control and HBPC Groups on Categorical Variables With Bootstrapping
      Control, % (n)HBPC-Receiving, % (n)χ2,
      After adjusting for statistical bias via bootstrapping based on 1000 samples.
      Two-Sided P
      Sex
       Females39.8 (219)42.0 (87)0.58.62
       Males60.2 (331)58.0 (120)
      Primary diagnostic category
       Mild ID41.6 (229)40.6 (84)1.93.59
       Moderate ID27.3 (150)31.9 (66)
       Severe ID12.9 (71)10.6 (22)
       Profound ID18.2 (100)16.9 (35)
      After adjusting for statistical bias via bootstrapping based on 1000 samples.
      Table 2Independent Sample t Tests Between the Control and HBPC Groups on Continuous Variables With Bootstrapping
      ControlHBPC-Receiving95% CI for Mean Difference
      After adjusting for statistical bias via bootstrapping based on 1000 samples.
      Two-Sided PGlass Δ
      Age, mean (SD)49.12 (18.73)48.99 (18.76)−3.17, 3.39.950.01
      Managed days, mean (SD)307.56 (106.56)317.56 (191.34)−41.68, 19.27.53−0.05
      Hospitalization rate per patient per year in the year prior to program, mean (SD)1.03 (3.66)0.32 (1.27)0.35, 1.08<.010.55
      Hospitalization rate per patient per year during program, mean (SD)
      Unadjusted for covariates.
      0.99 (2.94)0.19 (0.60)0.54, 1.07<.011.33
      After adjusting for statistical bias via bootstrapping based on 1000 samples.
      Unadjusted for covariates.
      The distribution of our dependent variable (ie, individual hospitalization rate per year) was significantly (P < .01) positively skewed (skewness = 7.15, SE = 0.089), which is fairly common for hospitalization data.
      • Williford E.
      • Haley V.
      • McNutt L.A.
      • Lazariu V.
      Dealing with highly skewed hospital length of stay distributions: the use of Gamma mixture models to study delivery hospitalizations.
      For these types of variables, gamma distribution is thought to be most appropriate, as it has shown in generalized linear models to produce parameter estimates with the closest data fit.
      • Ramos P.L.
      • Nascimento D.C.
      • Ferreira P.H.
      • Weber K.T.
      • Santos T.E.
      • Louzada F.
      Modeling traumatic brain injury lifetime data: Improved estimators for the Generalized Gamma distribution under small samples.
      Thus, in our generalized estimating equations analysis, the dependent variable was specified for gamma distribution with natural log (ln) link function and maximum likelihood parameter estimation method. Beta (β) regression coefficients of parameter estimates were transformed into odds ratios by taking EXP [regression coefficient] or eX.
      The results showed that group membership had a significant effect on the individual hospitalization rate (Wald χ2 = 20.71, P < .01) even after controlling for patients’ age and individual hospitalization rate in the year prior to the beginning of the study. Neither of the covariates was statistically significant. There was also no effect of sex on individual hospitalization rate (P = .22). The main finding of this study was that being in the control group was associated with a 2.12-fold increase in risk of hospitalization. These results are summarized in Table 3. Indeed, the mean hospitalization rate in the HBPC group during the study period was 0.19 (SD = 0.60) compared with 0.99 (SD = 2.94) in the control group.
      Table 3Parameter Estimates for Fixed Effects and Covariates in Generalized Estimating Equations Predicting Individual Hospitalization Rate
      ParameterβSE95% Wald CIOdds RatioWald χ2P
      Female−0.250.20−0.65, 0.150.781.53.22
      Control group0.750.170.42, 1.082.1220.71<.010
      Age0.0030.005−0.07, 0.011.000.29.59
      Hospitalization rate for the year prior to study0.030.03−0.02, 0.081.031.17.28

      Discussion

      This study reports hospitalizations in a population of medically complex individuals with IDD living in residential care settings in Ohio. We compared the hospitalization rate in individuals with IDD receiving HBPC to a similar population that did not receive HBPC, and we found that individuals receiving HBPC were hospitalized at a lower rate than those not receiving HBPC. An unexpected finding was that, in the year prior to the study period, the HBPC group had a significantly lower hospitalization rate than controls. Before we performed this analysis, we hypothesized that the selection of HBPC by a patient or responsible party may have been biased toward individuals with a higher baseline hospitalization rate opting out of traditional primary care, and opting for HBPC. What we found was the opposite. In fact, at baseline, patients who eventually opted for HBPC had approximately one-third of the hospitalization rate of the group receiving HBPC. What caused this baseline difference is uncertain, but potential factors include the following: patients or responsible parties and care team members may have already implemented care plans focused on reducing avoidable hospital admissions, they may have been cared for by primary care providers who were more responsive and engaged than those of the control group, or other congregate living setting–specific factors. These factors require additional study. Despite this baseline difference, we found that patients in the control group were more than 2 times more likely to be hospitalized than patients in the HBPC group, after controlling for the baseline difference in hospitalization rate in the year prior to the beginning of the study (as well as controlling for age). Contributors to the HBPC group's lower hospitalization rate during the study period may have included: availability of 24 × 7 medical triage for all HBPC patients, early recognition and management of new acute diseases and exacerbation of chronic conditions at frequent visits, a heightened focus on goal-directed care and advance care planning owing to the intrenchment of these facets in the group's care model, and improved management of the patient as a result of more intimate knowledge of the care setting in which the patient resides (by the practitioner visiting the home). More frequent home visits in the patient's preferred environment often lead to strong relationships between the HBPC practice and the community living staff, which may result in more effective between-visit care management and medical triage.
      Reporting of value-based outcome measures such as hospitalization rate in the management of individuals with IDD is increasingly important to patients, families, advocates, providers, payors, and state and federal policy makers. Although hospitalization can be distressing for most people, individuals with IDD can find the experience even more challenging, making this a particularly important quality measure for the population. The HBPC model is typically a more intensive community-based medical model, and one that may be ideally suited to medically complex populations of individuals with IDD who live in residential settings. HBPC providers partner with the staff serving the individual with IDD, using their shared knowledge of the patient's home setting to provide a more comprehensive assessment and more effective treatment plan. As recently discussed in JAMDA, there exists a considerable current opportunity to reimagine long-term care.
      • Zimmerman S.
      • Cesari M.
      • Gaugler J.E.
      • et al.
      The inevitability of reimagining long-term care.
      New, augmented models of support and care for people with IDD are needed, and HBPC is one such promising model of care. Reporting hospitalization rates among individuals with IDD, along with potential strategies to reduce them, may help to align individuals receiving services, families, provider organizations, and payors around the goal of enabling people with IDD to spend more time integrated in the community and less time in institutional settings.
      This study has limitations. Our population was focused on 1 state, individuals resided in congregate living settings managed by 1 organization, and patients received HBPC from 1 practice. An unexpected finding was that the HBPC had a lower hospitalization rate at baseline. However, we controlled for this difference in our statistical model, and during the study period, the HBPC group was hospitalized at a lower rate than the control group. Larger, longer-term studies are needed to further elucidate the impact HBPC has on these, and other, outcome measures in people with IDD.

      Conclusion and Implication

      In conclusion, we found that individuals with IDD receiving HBPC were hospitalized at a lower rate than a control group receiving traditional primary care. Expanding access to HBPC may be a worthwhile priority for organizations that support individuals with IDD.

      Acknowledgments

      The authors wish to express our most sincere gratitude to the direct support personnel, caregivers, nurses, leaders and advocates of programs, homes, and communities that provide support and care for people with IDD.

      References

        • Desai N.R.
        • Smith K.L.
        • Boal J.
        The positive financial contribution of home-based primary care programs: the case of the Mount Sinai Visiting Doctors.
        J Am Geriatr Soc. 2008; 56: 744-749
        • Wajnberg A.
        • Wang K.H.
        • Aniff M.
        • et al.
        Hospitalizations and skilled nursing facility placements before and after the implementation of a home-based primary care program.
        J Am Geriatr Soc. 2010; 58: 1144-1147
        • Hughes S.L.
        • Weaver F.M.
        • Giobbie-Hurder A.
        • et al.
        Effectiveness of team-managed home-based primary care.
        JAMA. 2000; 284: 2877-2885
        • Rotenberg J.
        • Kinosian B.
        • Boling P.
        • Taler G.
        Independence at Home Learning Collaborative Writing Group. Home-based primary care: beyond extension of the independence at home demonstration.
        J Am Geriatr Soc. 2018; 66: 812-817
        • Sloane P.D.
        • Zimmerman S.
        • Perez R.
        • et al.
        Physician perspectives on medical care delivery in assisted living.
        J Am Geriatr Soc. 2011; 59: 2326-2331
        • Mills W.R.
        • Robertson S.P.
        • Ogard K.L.
        • et al.
        Hospital care of assisted living residents by an assisted living based primary care team.
        Home Health Manag Pract. 2013; 25: 110-114
        • United Healthcare
        Quality improvement for individuals with intellectual and developmental disabilities.
        2016
        • Durbin A.
        • Balogh R.
        • Lin E.
        • et al.
        Repeat emergency department visits for individuals with intellectual and developmental disabilities and psychiatric disorders.
        Am J Intellect Dev Disabil. 2019; 124: 206-219
        • Durbin A.
        • Balogh R.
        • Lin E.
        • Wilton A.S.
        • Lunsky Y.
        Emergency department use: common presenting issues and continuity of care for individuals with and without intellectual and developmental disabilities.
        J Autism Dev Disord. 2018; 48: 3542-3550
        • O’Leary L.
        • Cooper S.A.
        • Hughes-McCormack L.
        Early death and causes of death of people with intellectual disabilities: a systematic review.
        J Appl Res Intellect Disabil. 2018; 31: 325-342
        • Shooshtari S.
        • Ouellette-Kuntz H.
        • Balogh R.
        • et al.
        Patterns of mortality among adults with intellectual and developmental disabilities in the Canadian province of Manitoba.
        J Policy Pract Intellect Disabil. 2020; 17: 270-278
        • Davies S.
        • Schmidt E.
        • Schultz E.
        • et al.
        Home and community-based services quality indicators.
        2010
        • Friedman C.
        Building the Framework for IDD Quality Measures.
        The Council on Quality and Leadership, the Institute for Public Policy for People with Disabilities, and Mosaic, 2018
        • Mills W.R.
        • Law K.
        • Boyd R.
        • et al.
        Hospitalization rate and days spent at home as value-based outcome measures for people with intellectual and developmental disability.
        Health Sci J. 2020; 14: 717
        • McCall N.
        • Geonnotti K.
        Utilization measures worksheet. Math Policy Res.
        • Twisk J.W.
        Applied Longitudinal Data Analysis for Epidemiology: A Practical Guide.
        2nd ed. Cambridge University Press, 2013
        • Ferguson C.J.
        An effect size primer: A guide for clinicians and researchers.
        in: Kazdin A.E. Methodological Issues and Strategies in Clinical Research. American Psychological Association, 2016: 301-310
        • Williford E.
        • Haley V.
        • McNutt L.A.
        • Lazariu V.
        Dealing with highly skewed hospital length of stay distributions: the use of Gamma mixture models to study delivery hospitalizations.
        PloS One. 2020; 15: e0231825
        • Ramos P.L.
        • Nascimento D.C.
        • Ferreira P.H.
        • Weber K.T.
        • Santos T.E.
        • Louzada F.
        Modeling traumatic brain injury lifetime data: Improved estimators for the Generalized Gamma distribution under small samples.
        PLoS One. 2019; 14: e0221332
        • Zimmerman S.
        • Cesari M.
        • Gaugler J.E.
        • et al.
        The inevitability of reimagining long-term care.
        J Am Med Dir Assoc. 2022; 23: 187-189