Introduction Access to care is essential for patients with diabetes to maintain health and prevent complications, and is important for health equity. New York State’s Health Homes (HHs) provide care management services to Medicaid-insured patients with chronic conditions, including diabetes, and aim to improve quality of care and outcomes. There is inconsistent evidence on the impact of HHs, and care management programs more broadly, on access to care.
Research design and methods Using a cohort of patients with diabetes derived from electronic health records from the INSIGHT Clinical Research Network, we analyzed Medicaid data for HH enrollees and a matched comparison group of HH non-enrollees. We estimated HH impacts on several access measures using natural experiment methods.
Results We identified and matched 11 646 HH enrollees; patients were largely non-Hispanic Black (29.9%) and Hispanic (48.7%), and had high rates of dual eligibility (33.0%), Supplemental Security Income disability enrollment (49.1%), and multiple comorbidities. In the 12 months following HH enrollment, HH enrollees had one more month of Medicaid coverage (p<0.001) and 4.6 more outpatient visits than expected (p<0.001, evenly distributed between primary and specialty care). There were also positive impacts on the proportions of patients with follow-up visits within 7 days (4 percentage points (pp), p<0.001) and 30 days (6pp, p<0.001) after inpatient care, and on the proportion of patients with follow-up visits within 30 days after emergency department (ED) care (4pp, p<0.001). We did not find meaningful differences in continuity of care. We found small positive impacts on the proportion of patients with an inpatient visit and the proportion with an ED visit.
Conclusions New York State’s HH program improved access to care for Medicaid recipients with diabetes. These findings have implications for New York State Medicaid as well as other providers and care management programs.
- diabetes mellitus
- type 2
Data availability statement
Data may be obtained from a third party and are not publicly available. The data that support the findings of this study are not publicly available.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Significance of this study
What is already known about this subject?
Care management programs have not been found to impact hospital admissions, but there is inconsistent evidence of an effect on primary care (outpatient) services or other measures of access.
What are the new findings?
New York State Medicaid’s Health Home program was associated with significant increases in access to care among patients with diabetes:
In the 12 months following HH enrollment, HH enrollees had 1 more month of Medicaid coverage (p<0.001) than expected.
In the 12 months following HH enrollment, HH enrollees had 4.6 more outpatient visits than expected (p<0.001, evenly distributed between primary and specialty care).
There were also positive impacts on the proportions of patients with follow-up visits within 7 days (4 percentage points (pp), p<0.001) and 30 days (6pp, p<0.001) after inpatient care, and on the proportion of patients with follow-up visits within 30 days after emergency department care (4pp, p < 0.001).
How might these results change the focus of research or clinical practice?
These findings can inform future research that examines the impact of care management programs on access to care, a rarely studied, but patient-centered set of outcomes relevant to achieving health equity.
In addition, the findings suggest future analyses should examine the clinical significance of observed differences in access to care measures.
Access to care is an important determinant of health, disease status, and equity.1 2 Access to specialty care visits, along with primary care visits, has been linked to greater receipt of guideline concordant care and better functional outcomes.3 4 Some evidence indicates that access to prompt (within 7 days) postdischarge follow-up primary care visits have been associated with fewer 30-day readmissions to the hospital.5–8 Continuity of primary care, an aspect of access to care, has been associated with fewer hospitalizations and emergency room visits, lower costs, better patient satisfaction, and lower mortality.9–15 Due to systemic inequity in the USA, Black and Latinx individuals, people with low income, and those with certain chronic illnesses and disabilities have inadequate access to care.2 16 17
For patients with diabetes, having health insurance coverage and adequate access to necessary and timely health services is essential for managing their illness and preventing complications.18–20 Access has been associated with quality and outcomes of care among patients with diabetes including glycemic and other risk factor control, rates of screenings for kidney, eye, and foot complications, prompt care for acute and chronic illnesses, timely specialty referrals, and complications.3 21–26
The principal goal of care management programs is to improve access to and coordination of care. As part of the Patient Protection and Affordable Care Act of 2010, ‘health homes’ (HHs) were authorized and funded to manage and coordinate care for individuals enrolled in Medicaid with complex chronic conditions.27 Diabetes was one of the conditions prioritized for HH enrollment. Six services are eligible for reimbursement through the HH program: comprehensive care management, care coordination and health promotion, comprehensive transitional care assistance, patient and family support, referral to community and social support services, and use of health information technology to link services.28 New York State established its Medicaid Health Home program in 2011; eligible individuals include people living with HIV, those with serious mental illness, or those with two or more chronic conditions.29 Through coordination of care, intensive case management, and attention to social needs, the program intends to help individuals with diabetes and other conditions establish more steady relationships with primary healthcare teams, gain greater access to outpatient services, and improve transitions between care sites. Qualitative information collected previously from our study indicates that HH members view assistance with accessing care to be one of the most desired and valuable services provided by their care managers.30
While improving access to care is recognized and important, especially for underserved populations and for those with diabetes, the evidence of a beneficial impact of care management programs is sparse, especially among those with complex needs. The implementation of the HH program in some, but not all eligible patients with diabetes, presents a ‘natural experiment’ through which to study these impacts, as a goal of the HH program is to assure that patients get the care they need when they need it. Therefore, in this analysis, we sought to determine the effect of HH enrollment on access to care for patients with diabetes and other chronic conditions by examining the number of outpatient visits (both primary care and specialty care), ED visits, and hospitalizations, transitions in care, continuity of care (COC), and Medicaid coverage in the 12 months prior to enrollment compared with the 12 months post enrollment in HH members compared with a matched comparison group of Medicaid-insured patients with diabetes who did not enroll in the HH.
Study population, data sources, and setting
We derived our sample from the INSIGHT Clinical Research Network electronic health record (EHR) data from six large healthcare systems in New York City.31 The cohort included individuals who met modified SUrveillance, PREvention, and ManagemEment of Diabetes Mellitus (SUPREME-DM)32 criteria for diabetes (diagnosis, laboratory, and/or medication criteria) between 2010 and 2016, were insured by Medicaid, and were over 18 years of age as of January 1, 2010. Additional Medicaid HH program data were then used to identify HH enrollees and their periods of enrollment. Access to care data for this study was derived from fee-for-service claims and managed care encounter records from the New York State Department of Health for calendar years 2010–2017.
Individuals need to consent to enroll in an HH; they are not automatically a part of the program, even if they are eligible. HH programs received reimbursement for outreach to eligible individuals for up to a 3-month period prior to enrollment, though some HH participants enrolled directly. We defined our HH enrollee group (‘program group’) as any individual with at least 1 month of HH enrollment (n=11 646). Our matched comparison group included individuals who never enrolled in an HH; this includes individuals who received outreach but did not ultimately enroll (n=10 640), as well as individuals who did not receive any outreach (n=15 961). This does not mean, however, that outreached individuals declined HH services. In fact, non-enrollment despite outreach was frequently attributed to difficulties in contacting individuals, often due to a lack of up-to-date contact information.28
We defined a 12-month ‘baseline’ period immediately predating, and a 12-month ‘follow-up’ period immediately postdating, an ‘index HH activation date’. For the program group, this index date was the earliest date of outreach by or enrollment in an HH. For comparison group members in the outreach group who never enrolled in an HH, the index date was the earliest date of outreach. For each potential comparison group member who received neither outreach nor enrollment, we randomly selected one index date in 2011–2016, with a probability corresponding to the frequency distribution of the number of Medicaid enrolled months in the baseline periods of program group members. For example, if 90% of program group members were enrolled in Medicaid in all 12 baseline period months, comparison group index dates immediately preceded by 12 enrolled months in Medicaid had a 90% probability of being selected. We then conducted 5:1 matching with replacement using both the outreached but never enrolled group and the selected dates from the never outreached or enrolled group (see below for details on matching methods).
As Medicaid data were available for calendar years 2010–2017, Medicaid recipients whose HH index date occurred in calendar years 2011–2016 and who were subsequently enrolled were included in the program group to allow for complete observation of Medicaid utilization for 1 year before, and 1 year after, their index date.
Several sociodemographic and health services characteristics including age, gender, race/ethnicity, receipt of Medicare (dual-eligible status), receipt of Supplemental Security Income (SSI), and Medicaid enrollment in each calendar month were directly derived from Medicaid enrollment and eligibility data. Care utilization was measured in separate categories of services using procedure and rate codes, place and category of service, and provider specialty on claim (in the case of fee-for-service) or encounter (in the case of managed care) records. Diagnoses on claims or encounters were used to create indicators for the presence of several disease conditions in the baseline period.
We used the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project (HCUP) Clinical Classification Software (CCS) programs applied to International Classification of Diseases (ICD) diagnoses on claims to create indicators of any cancer (CCS 11–36, exclude 22), serious mental illness (CCS 657, 658, 659, 662), substance use disorders (CCS 660, 661), HIV (CCS 5, 9), and diabetes (CCS 49, 50).33 Diabetes diagnoses were further categorized into separate indicators for a diagnosis received in an inpatient setting or an outpatient setting. A count of chronic conditions was created using the HCUP chronic condition indicator (CCI).34
Care access measures included:
Outpatient utilization: Counts of primary care visits (defined as office visits with primary care, internal medicine, or family practice provider specialties) and specialty care visits (defined as all other specialties except for psychiatry), and total outpatient visits as the sum of the two.
Medicaid enrollment: The number of months enrolled in Medicaid.
Transitions in care measures (recorded as missing value for individuals without an inpatient stay, or without an emergency department (ED) visit, in the relevant (baseline or follow-up) period), as follows:
The proportion of an individual’s inpatient stays followed by an outpatient visit within 7 days, and within 30 days, of discharge.
The proportion of an individual’s ED visits followed by an outpatient visit within 7 days, and within 30 days, of the ED visit.
Continuity of care:
The Bice-Boxerman COC Index for the baseline or follow-up period, which is a claims-based measure of the degree to which a patient’s visits over a time period or illness are dispersed among a single or multiple providers.35 36
The Usual Provider of Care (UPC) Index for the baseline or follow-up period, which is a claims-based measure of the concentration of care with a single, primary provider.35
The proportion of individuals with at least one inpatient stay.
The proportion of individuals with at least one ED visit.
Matching and weighting
Our matching procedure first identified the five ‘nearest neighbors’ for each program group member ‘with replacement’ using Mahalanobis distance as the distance metric.37 An entropy balance weight was then calculated in order to estimate the average effect of treatment on the treated. Entropy balance weights generalize the ‘synthetic control method’ to include multiple treated cases, and have been shown to achieve near perfect covariate balance.38
Matching variables included all demographic and baseline period utilization measures described in the ‘Measures’ section above, as well as total Medicaid payments recorded in the claims and encounter data. In order to match on the utilization time trend, total Medicaid cost was matched for each calendar month in the baseline period. This combination of matching on time invariant characteristics such as demographics, and matching on repeated measures that capture the time trend, reduces confounding introduced by ‘differential mean reversion’ between the treated and comparison groups, by which an estimated treatment ‘effect’ may be an artifact of different rates of regression to the mean due to unobserved differences between the groups.39
The COC indices (UPC and COC) were also measured in the baseline period and used as matching variables, but because UPC is undefined for individuals with no visits, and COC is undefined for individuals with fewer than two visits, these measures were set to zero when undefined, and a missing indicator for each measure was included in the matching procedure, in order to retain individuals with few or no outpatient visits in the baseline period.
In order to estimate the HH program’s impact on access and COC in the 12-month follow-up period, regression models were estimated of the form:
Outcomeit = β0 + β1Enrolledi + β2Outcomei(t-1) + γXi +ei, where Outcomei(t-1) represents the level of the outcome during the baseline period, and Xi represents a vector of covariates including additional baseline inpatient and outpatient utilization measures and sociodemographic characteristics. Additional impact models were estimated while also adjusting for the number of months of Medicaid enrollment in the follow-up period, in order to determine the extent to which estimated program effects were dependent on the program’s effect on Medicaid enrollment. Because maintaining Medicaid enrollment can be an important tool for maintaining access to care, our main results do not adjust for Medicaid enrollment in the follow-up period, but we include enrollment-adjusted results in an online supplemental table. Enrolledi is an indicator of HH enrollment, thus the main quantity of interest is β1, which represents the estimated impact of enrollment on the outcome. Including a lagged outcome measured in the baseline period as a regressor enables us to control for the comparison group pre-to-post change within a difference-in-differences framework. All models used entropy balance weights as probability weights, and robust standard errors were estimated.
Fractional logistic regression models were estimated for outcomes whose values ranged from 0 to 1 (COC and UPC measures, and the proportion of hospitalizations or ED visits followed up within 7, or 30, days by an outpatient visit). Logistic regression models were estimated for the presence of any hospitalization or any ED visit in the follow-up period. Negative binomial models were estimated for the number of primary care, specialty care, and total outpatient (primary care plus specialty care) visits.
Because the 7/30-day follow-up measures are only defined for individuals experiencing an ‘index’ event (either a hospitalization or and ED visit), and the COC and UPC measures are only defined for individuals having at least two, or at least one, outpatient visit, respectively, these models were estimated only on individuals meeting these criteria in the follow-up period. In order to retain all such individuals in the models, missing baseline COC and UPC indices and follow-up measures for the baseline period were set to zero and missing indicators were included for each of these when used as baseline period covariates.
Since the COC and UPC indices are difficult to interpret, Cohen’s d effect sizes were calculated for all impacts.40 For logistic regression results, log-odds were converted to Cohen’s d using the formula: d=log(OR) * sqrt(3) / pi.41
We identified 115 434 individuals with diabetes using the INSIGHT EHR data. Table 1 presents the baseline characteristics for the 11 646 individuals who enrolled in an HH in 2010–2016. A sizeable percentage of HH enrollees were non-Hispanic Black (29.9%) or Hispanic (48.7%) individuals; 33.0% were dual-eligible, and 49.1% received SSI disability. As expected, this population had complex needs indicated by multiple comorbidities, including 53.7% with serious mental illness, 26.9% with alcohol and substance use disorders, and a mean of 17.0 CCI conditions. In the baseline period, the group had high rates of healthcare utilization (mean of 1.0 hospitalization, 1.1 ED visits) and Medicaid payments (mean $45 488 in nominal dollars).
Tables 2 and 3 display the results of our matching and weighting procedures. After matching and weighting, standardized differences between the mean values for the matched groups for all matching variables were 0, and variance ratios were 1 (within measurement error). All HH enrollees in the unmatched sample were retained in the matched sample.
Our analyses estimated HH program impacts as the difference between the observed mean level of each outcome and what would have been expected based on observed levels of the outcome variable in the comparison group, adjusting for baseline covariates that included observed levels of the baseline outcome in each group (table 4). In the 12 months following HH enrollment, HH enrollees had, on average, 1.0 more month of Medicaid coverage (p<0.001) and 4.6 more outpatient visits than expected (p<0.001). The program impact on outpatient visits was evenly distributed between primary and specialty care visits. There were also positive impacts on the proportions of patients with follow-up visits within 7 days (4 percentage points (pp), p<0.001) and 30 days (6pp, p<0.001) after an inpatient admission, and on the proportion of patients with follow-up visits within 30 days after an ED visit (4pp, p<0.001), and a marginally statistically significant positive impact on follow-up visits within 7 days after an ED visit (2pp, p=0.056). There was a small, negative program impact on the UPC Index (−0.02, p<0.001), and no detectable effect on the COC Index. There were also positive impacts on the proportions of patients with an inpatient stay (7pp, p<0.001) or an ED visit (7pp, p<0.001), largely attributable to decreases in these measures in the comparison group. When adjusting for months enrolled in Medicaid in the follow-up period, we found smaller impacts on outpatient visits that remained statistically significant, smaller impacts of follow-up visits within 30 days of an inpatient admission or ED visit that remained statistically significant and similar findings with regard to UPC and COC indices (online supplemental table 1).
We examined access to care among Medicaid Health Home participants with diabetes before and after enrollment compared with a matched and weighted comparison group. Overall, being in an HH was associated with significant improvements in access to care. We found that HH enrollees had more outpatient visits (both primary care and specialty visits), prompter follow-up after ED visits and admissions, and more months of Medicaid coverage in the 12-month postenrollment period. Each of these results is plausibly clinically significant (several more outpatient visits, 2–6pp increases in prompt follow-up, and one additional month of Medicaid-insurance coverage). We found a very small negative impact on one COC measure, however, for HH enrollees compared with non-enrollees; given the small effect size (d=−0.073) this effect does not seem likely to be clinically significant. We also found a higher percentage of patients with inpatient and ED visits than expected. We should note that our impact estimates were attenuated toward zero when adjusting for the increase in Medicaid coverage in the follow-up period (see online supplemental table 1), such that about half of the program’s impacts on outpatient, primary, and specialty care appeared to be due to the program’s impact on Medicaid enrollment, but these categories of service increased even after adjustment for longer Medicaid enrollment periods in the HH group.
Our study adds evidence about the impact of care management programs generally. While a principal goal of most care management programs is to reduce inpatient and ED utilization, and therefore costs, multiple studies, meta-analyses, and systematic reviews have found mixed effects of care management programs on healthcare utilization.42–49 An evaluation of the Chronic Illness Demonstration Project, the forerunner of the HH program in New York State, found a similar increase in outpatient visits.50 Most recently, Buja et al noted consistent findings across meta-analyses and systemic reviews of no impact on hospital admissions, and variable evidence of impact on ED visits, costs, length of stay, and primary care.49 But evaluations focused only on inpatient and ED utilization and costs may not capture all important impacts, especially from the patient perspective. A relatively small number of studies include outpatient care as an outcome, and these rarely evaluate other measures of access to care. In addition, very little of this research focuses on patients with complex needs.48 Therefore, our study adds to two important areas currently lacking robust evidence, namely access to care and individuals with complex needs.
Poor access to care is one reason for known disparities in diabetes care and outcomes experienced by Black and Latinx populations and those with serious mental illness.26 51–54 Our study contributes by estimating the impact of a policy among a group largely comprised of individuals at risk for poor access because of systemic inequity (the group was 29.9% non-Hispanic Black individuals and 48.7% Hispanic individuals, and 53.7% had a diagnosis of serious mental illness). This study also echoes the findings of our qualitative work, which found that HH improved access to care from the patient perspective.30
Our study has several limitations. We examined a set of intermediate markers of access to care and not clinical outcomes among patients with diabetes. Future analyses will examine whether the HH improved clinical outcomes including the receipt of guideline concordant care, better disease control, and fewer diabetes-related hospitalizations. The cohort we studied is a subset of patients in New York City (NYC) with diabetes who are insured by Medicaid, since it is limited to those who received at least some of their care at one of the academic medical centers that participate in the INSIGHT Clinical Research Network. As such, our sample does not represent the experience of all Medicaid-insured patients with diabetes, as patients who receive care only at NYC Health+Hospitals locations (NYC’s public healthcare system) or community hospitals are not included in our cohort.
In addition, although we used advanced observational research natural experiment methods to identify and weight the comparison group, it is still possible that there were differences between the groups that we did not account for, leading to biased results. Specifically, it is possible that there were remaining unobserved differences between the groups that created ‘differential mean reversion’, by which individual outcome values regress to their respective population means at different rates in the two groups because the groups are drawn from fundamentally different underlying populations.39 We should note, for example, that most of our comparison group’s average outcome values decrease from the baseline to the follow-up period, whereas all of the HH enrolled group’s outcome values increase from baseline to follow-up. This suggests the possibility that our estimated program impacts may be biased by unobserved or unadjusted differences in illness severity that would have resulted in shorter episodes of care in the comparison cohort than in the HH enrolled group even in the absence of the HH program. As we described above, we have addressed this possibility by matching on time trends in utilization, as well as on time invariant characteristics.
This study examined data from 2010 to 2017, prior to the global pandemic of COVID-19. COVID-19 severely disrupted healthcare delivery, and thus access to care, in NYC and throughout the USA. The individuals who are members of the HH program are at particularly high risk for the morbidity, mortality, and care disruption caused by COVID-19. This study serves as a baseline description of access to care in a particularly vulnerable population. It suggests that for patients with diabetes and Medicaid insurance, care management services provided by the HH could serve an important role in connecting patients with providers and services; this connection may be even more important under pandemic circumstances, and will be a focus of our future research.
In conclusion, our results suggest that the HH program had a positive and favorable impact on access to care among patients with diabetes in NYC during the program’s initial 5 years. These findings add to understanding the potential role that care management programs have in improving care.
Data availability statement
Data may be obtained from a third party and are not publicly available. The data that support the findings of this study are not publicly available.
Patient consent for publication
This study was approved by the Biomedical Research Alliance of New York’s Institutional Review Board (BRANY, # 16-08-299-05) with a full waiver of the requirement to obtain authorization for use or disclosure of protected health information.
The authors thank the study’s Patient/Stakeholder Advisory Board for their essential role in this work. Also, the authors would like to thank the Natural Experiments for Translation in Diabetes (NEXT-D2) research network for invaluable support and guidance throughout this study.
Contributors VM and DS, the Principal Investigators of the study, worked with TM, JF, AM, NE, and RD to conceive of and design the study. TM developed the methods and conducted analyses, with input from VM, DS, JF, AM, NE, RD, and MC. All authors contributed to interpretation of results and implications. VM and DS led manuscript preparation. VM and TM drafted the manuscript; DS, JF, AM, NE, RD, and MC reviewed the manuscript and contributed to substantial modifications.
Funding Research reported in this manuscript was funded through a Patient-Centered Outcomes Research Institute (PCORI) Award (NEN-1508–32252), and by the PCORI Award CER-2017C3-9230 and the National Cancer Institute (P30-CA008748). This work was supported by the INSIGHT Clinical Research Network, funded by the Patient-Centered Outcomes Research Institute Award, RI-CRN-2020–004 (Principal Investigator: Rainu Kaushal, MD, MPH).
Disclaimer The views and opinions expressed in this report are those of the authors and do not necessarily reflect the official policy or position of the New York State Department of Health. Examples of analysis performed within this report are only examples. They should not be utilized in real-world analytic products.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.