Issues and biases in matching medicaid pregnancy episodes to vital records data: the Arkansas experience

Case study on communicating with research ethics committees about minimizing risk through software: an application for record linkage in secondary data analysis

Objective

In retrospective secondary data analysis studies, researchers often seek waiver of consent from institutional Review Boards (IRB) and minimize risk by utilizing complex software. Yet, little is known about the perspectives of IRB experts on these approaches. To facilitate effective communication about risk mitigation strategies using software, we conducted two studies with IRB experts to co-create appropriate language when describing a software to IRBs.

Materials and Methods

We conducted structured focus groups with IRB experts to solicit ideas on questions regarding benefits, risks, and informational needs. Based on these results, we developed a template IRB application and template responses for a generic study using privacy-enhancing software. We then conducted a three-round Delphi study to refine the template IRB application and the template responses based on expert panel feedback. To facilitate participants' deliberation, we shared the revisions and a summary of participants' feedback during each Delphi round.

Results

11 experts in two focus groups generated 13 ideas on risks, benefits, and informational needs. 17 experts participated in the Delphi study with 13 completing all rounds. Most agreed that privacy-enhancing software will minimize risk, but regardless all secondary data studies have an inherent risk of unexpected disclosures. The majority (84.6%) noted that subjects in retrospective secondary data studies experience no greater risks than the risks experienced in ordinary life in the modern digital society. Hence, all retrospective data-only studies with no contact with subjects would be minimal risk studies.

Conclusion

First, we found fundamental disagreements in how some IRB experts view risks in secondary data research. Such disagreements are consequential because they can affect determination outcomes and might suggest IRBs at different institutions might come to different conclusions regarding similar study protocols. Second, the highest ranked risks and benefits of privacy-enhancing software in our study were societal rather than individual. The highest ranked benefits were facilitating more research and promoting responsible data governance practices. The highest ranked risks were risk of invalid results from systematic user error or erroneous algorithms. These societal considerations are typically more characteristic of public health ethics as opposed to the bioethical approach of research ethics, possibly reflecting the difficulty applying a bioethical approach (eg, informed consent) in secondary data studies. Finally, the development of privacy-enhancing technology for secondary data research depends on effective communication and collaboration between the privacy experts and technology developers. Privacy is a complex issue that requires a holistic approach that is best addressed through privacy-by-design principles. Privacy expert participation is important yet often neglected in this design process. This study suggests best practice strategies for engaging the privacy community through co-developing companion documents for software through participatory design to facilitate transparency and communication. In this case study, the final template IRB application and responses we released with the open-source software can be easily adapted by researchers to better communicate with their IRB when using the software. This can help increase responsible data governance practices when many software developers are not research ethics experts.

Maternal comorbidity index and severe maternal morbidity among medicaid covered pregnant women in a US Southern rural state

BACKGROUND

The rates of SMM have been steadily increasing in Arkansas, a southern rural state, which has the 5th highest maternal death rate among the US states. The aims of the study were to test the functionality of the Bateman index in association to SMM, in clustering the risks of pregnancies to SMM, and to study the predictability of SMM using the Bateman index.

STUDY DESIGN

From the ANGELS database, 72,183 pregnancies covered by Medicaid in Arkansas between 2013 and 2016 were included in this study. The expanded CDC ICD-9/ICD-10 criteria were used to identify SMM. The Bateman comorbidity index was applied in quantifying the comorbidity burden for a pregnancy. Multivariable logistic regressions, KMeans method, and five widely used predictive models were applied respectively for each of the study aims.

RESULTS

SMM prevalence remained persistently high among Arkansas women covered by Medicaid (195 per 10,000 deliveries) during the study period. Using the Bateman comorbidity index score, the study population was divided into four groups, with a monotonically increasing odds of SMM from a lower score group to a higher score group. The association between the index score and the occurrence of SMM is confirmed with statistical significance: relative to Bateman score falling in 0-1, adjusted Odds Ratios and 95% CIs are: 2.1 (1.78, 2.46) for score in 2-5; 5.08 (3.81, 6.79) for score in 6-9; and 8.53 (4.57, 15.92) for score ≥10. Noticeably, more than one-third of SMM cases were detected from the studied pregnancies that did not have any of the comorbid conditions identified. In the prediction analyses, we observed minimal predictability of SMM using the comorbidity index: the calculated c-statistics ranged between 62% and 67%; the Precision-Recall AUC values are <7% for internal validation and <9% for external validation procedures.

CONCLUSIONS

The comorbidity index can be used in quantifying the risk of SMM and can help cluster the study population into risk tiers of SMM, especially in rural states where there are disproportionately higher rates of SMM; however, the predictive value of the comorbidity index for SMM is inappreciable.

Evaluation of a telemedicine program managing high-risk pregnant women with pre-existing diabetes in Arkansas's Medicaid program

OBJECTIVE

We aim to evaluate the effects of the telemedicine program, High-Risk Pregnancy Program at University of Arkansas for Medical Sciences (UAMS), on health services utilization and medical expenditures among pregnant women with pre-existing diabetes and their newborns.

RESEARCH DESIGN AND METHODS

The study sample was selected from the Arkansas Medicaid claims linked to infant birth/death certificates and UAMS telemedicine records from 2013 through 2016. We used propensity score matching based on participants' characteristics to create three groups - UAMS telemedicine care, UAMS in-person care, and non-UAMS prenatal care. We compared inpatient and outpatient care services, medication use and caesarean section rates, severe maternal morbidity, infant mortality and preterm birth rates and medical expenditures.

RESULTS

The UAMS telemedicine group had fewer inpatient admissions (1.18 vs 1.31; 95% CI: -0.27, 0.00), lower insulin use rates (41.86% vs 59.88%; 95% CI: -29.00%, -7.05%) and lower maternal care expenditures ($7,846 vs $10,644; 95% CI: -$4,089, -$1,507) compared with the UAMS in-person care group. Women receiving UAMS telemedicine had more prenatal care visits (10.45 vs 8.57; 95% CI: -2.96, -0.81), higher insulin use rates (41.86% vs 26.74%: 95% CI: 4.63%, 25.60%) and similar maternal care expenditures ($7,846 vs $7,051), compared with those receiving non-UAMS in-person care. Caesarean section, severe maternal morbidity, and infant mortality rates were similar across the three groups.

CONCLUSION

UAMS telemedicine was associated with improved utilization of prenatal care among pregnant women with pre-existing diabetes. Telemedicine services did not differ from usual in-person services in clinical outcomes and medical expenditures.

Improving perinatal regionalization: 10 years of experience with an Arkansas initiative

OBJECTIVE

To examine the impact of Antenatal and Neonatal Guidelines, Education and Learning Systems (ANGELS) on neonatal intensive care unit (NICU) preterm delivery rates.

STUDY DESIGN

In this longitudinal observational study, linked vital records and Medicaid claims records for 29,124 preterm births (April 2001-December 2012) to Medicaid covered women were used to examine factors predicting whether deliveries occurred at hospitals with neonatology-staffed NICUs. The factors associated with delivery are estimated and compared for baseline and three post-implementation periods.

RESULTS

Rates for NICU preterm deliveries increased from 28 to 37% over the time period. Compared to baseline, adjusted NICU delivery rates in the middle and late implementation periods were statistically significant (p < 0.001). Negative impacts of long travel times were reduced, while impacts of obstetrician prenatal care changed from negative to positive association.

CONCLUSION

Findings validate the ANGELS initiative premise: academic specialists, working with community-based care providers, can improve perinatal regionalization.

Ascertainment of medicaid payment for delivery on the iowa birth certificate: is accuracy sufficient for timely policy and program relevant analysis?

The Iowa Department of Public Health annually links Medicaid claims data to the birth certificate. Because the latest version of the birth certificate provides more timely and less costly information on delivery payment source, we were interested in assessing the validity and reliability of the birth certificate payment source compared to Medicaid paid claims. We linked Medicaid paid claims to birth certificates for calendar years 2007-2009 (n = 120,626). We measured reliability by Kappa statistic and validity by sensitivity, specificity, predictive value positive and negative. We examined reliability and validity overall and by maternal characteristics (e.g. age, race, ethnicity, education). The Kappa statistic for the birth certificate payment source indicated substantial agreement (0.78; 95 % CL 0.78-0.79). Sensitivity and specificity were also high, 86.3 % (95 % CL 86.0-86 6 %) and 91.9 % (95 % CL 91.7-92.1 %), respectively. The predictive value positive was 87.0 %. The predictive value negative was 91.4 %. Kappa and specificity were lower among women of racial and ethnic minorities, women younger than age 24, and women with less education. The overall Kappa, sensitivity and specificity generally suggest the birth certificate payment source is as valid and reliable as the linked data source. The birth certificate payment source is less valid and reliable for women of racial and ethnic minorities, women younger than age 24, and those with less education. Consequently caution should be exercised when using the birth certificate payment source for monitoring service use by the Medicaid population within specific population subgroups.

Considerations in using registry and health plan data for studying pregnancy in rheumatic diseases

PURPOSE OF REVIEW

The purpose of this review is to critically evaluate the strengths and limitations of different data sources for pregnancy-related research in patients with rheumatic diseases. We describe challenges in studying adverse pregnancy outcomes in the setting of observational research, with a particular focus on the studies of maternal drug exposures.

RECENT FINDINGS

We discuss potential threats to validity in the assessment of exposure and outcomes and controlling for confounding; present findings from selected pregnancy-related observational studies conducted using data from registries and health plans; and highlight future research opportunities for pregnancy research.

SUMMARY

Registry and health plan data contribute complementary information to each other. Used together, linked data sources may allow clinicians, researchers, and patients to obtain a more complete understanding of the risk and benefits associated with maternal drug exposure and of the risk factors associated with adverse birth outcomes in women with rheumatic disease.

Methodological challenges in using routinely collected health data to investigate long-term effects of medication use during pregnancy

To date, the investigation of teratogenic effects of medications has largely focused on physical alterations present at birth (i.e. malformations) as opposed to functional alterations (i.e. neurodevelopment, metabolic function) that may not be apparent at birth but could influence an individual's health and risk of disease in later life. The use of routinely collected health data represents one approach to better identifying, quantifying, and understanding the long-term risks or benefits of medication use during pregnancy. As such, the objective of this review was to identify and explore opportunities and challenges associated with using routinely collected health data to examine long-term effects of medication use during pregnancy. Drawing on published research several key methodological issues associated with their use in investigating long-term outcomes are reviewed. While significant opportunities exist to make greater use of routinely collected health data, there are a number of key challenges. Identified challenges relate to aspects of study design and analysis, and include obtaining access to data, the ability to match records across datasets and over long periods of time, how medication exposures are ascertained and classified, issues around loss to follow-up how outcomes are ascertained and classified, and the careful interpretation of results in light of study and data limitations. Understanding key challenges associated with using routinely collected health data to investigate long-term effects of medication use during pregnancy is essential in supporting their appropriate use and interpretation, which will contribute to improving the quality of research undertaken and ensure the reliability of results obtained.

Privacy preserving interactive record linkage (PPIRL)

OBJECTIVE

Record linkage to integrate uncoordinated databases is critical in biomedical research using Big Data. Balancing privacy protection against the need for high quality record linkage requires a human-machine hybrid system to safely manage uncertainty in the ever changing streams of chaotic Big Data.

METHODS

In the computer science literature, private record linkage is the most published area. It investigates how to apply a known linkage function safely when linking two tables. However, in practice, the linkage function is rarely known. Thus, there are many data linkage centers whose main role is to be the trusted third party to determine the linkage function manually and link data for research via a master population list for a designated region. Recently, a more flexible computerized third-party linkage platform, Secure Decoupled Linkage (SDLink), has been proposed based on: (1) decoupling data via encryption, (2) obfuscation via chaffing (adding fake data) and universe manipulation; and (3) minimum information disclosure via recoding.

RESULTS

We synthesize this literature to formalize a new framework for privacy preserving interactive record linkage (PPIRL) with tractable privacy and utility properties and then analyze the literature using this framework.

CONCLUSIONS

Human-based third-party linkage centers for privacy preserving record linkage are the accepted norm internationally. We find that a computer-based third-party platform that can precisely control the information disclosed at the micro level and allow frequent human interaction during the linkage process, is an effective human-machine hybrid system that significantly improves on the linkage center model both in terms of privacy and utility.

Telemedicine collaboration improves perinatal regionalization and lowers statewide infant mortality

OBJECTIVE

We assessed a telemedicine (TM) network's effects on decreasing deliveries of very low birth weight (VLBW, <1500 g) neonates in hospitals without Neonatal Intensive Care Units (NICUs) and statewide infant mortality.

STUDY DESIGN

This prospective study used obstetrical and neonatal interventions through TM consults, education and census rounds with 9 hospitals from 1 July 2009 to 31 March 2010. Using a generalized linear model, Medicaid data compared VLBW birth sites, mortality and morbidity before and after TM use. Arkansas Health Department data and χ(2) analysis were used to compare infant mortality.

RESULT

Deliveries of VLBW neonates in targeted hospitals decreased from 13.1 to 7.0% (P=0.0099); deliveries of VLBW neonates in remaining hospitals were unchanged. Mortality decreased in targeted hospitals (13.0% before TM and 6.7% after TM). Statewide infant mortality decreased from 8.5 to 7.0 per 1000 deliveries (P=0.043).

CONCLUSION

TM decreased deliveries of VLBW neonates in hospitals without NICUs and was associated with decreased statewide infant mortality.

Harnessing the Medicaid Analytic eXtract (MAX) to Evaluate Medications in Pregnancy: Design Considerations

BACKGROUND

In the absence of clinical trial data, large post-marketing observational studies are essential to evaluate the safety and effectiveness of medications during pregnancy. We identified a cohort of pregnancies ending in live birth within the 2000-2007 Medicaid Analytic eXtract (MAX). Herein, we provide a blueprint to guide investigators who wish to create similar cohorts from healthcare utilization data and we describe the limitations in detail.

METHODS

Among females ages 12-55, we identified pregnancies using delivery-related codes from healthcare utilization claims. We linked women with pregnancies to their offspring by state, Medicaid Case Number (family identifier) and delivery/birth dates. Then we removed inaccurate linkages and duplicate records and implemented cohort eligibility criteria (i.e., continuous and appropriate enrollment type, no private insurance, no restricted benefits) for claim information completeness.

RESULTS

From 13,460,273 deliveries and 22,408,810 child observations, 6,107,572 pregnancies ending in live birth were available after linkage, cleaning, and removal of duplicate records. The percentage of linked deliveries varied greatly by state, from 0 to 96%. The cohort size was reduced to 1,248,875 pregnancies after requiring maternal eligibility criteria throughout pregnancy and to 1,173,280 pregnancies after further applying infant eligibility criteria. Ninety-one percent of women were dispensed at least one medication during pregnancy.

CONCLUSIONS

Mother-infant linkage is feasible and yields a large pregnancy cohort, although the size decreases with increasing eligibility requirements. MAX is a useful resource for studying medications in pregnancy and a spectrum of maternal and infant outcomes within the indigent population of women and their infants enrolled in Medicaid. It may also be used to study maternal characteristics, the impact of Medicaid policy, and healthcare utilization during pregnancy. However, careful attention to the limitations of these data is necessary to reduce biases.

Medicaid and preterm birth and low birth weight: the last two decades

OBJECTIVES

To determine if (1) birth outcomes among women on Medicaid differ significantly from outcomes of those with private insurance, after controlling for known risk factors, and (2) enhanced prenatal care influences care use and birth outcomes.

METHODS

This is a review of studies published between 1989 and 2009 that examined birth outcomes (1) between women on Medicaid and those with private insurance and (2) among Medicaid enrollees who received comprehensive prenatal care.

RESULTS

When corrected for risk variables, birth outcomes are not different between private insurance and Medicaid patients. The impact of comprehensive prenatal care programs on birth outcomes varies across states and regions.

CONCLUSIONS

There is a need for critical evaluation of comprehensive programs in a regional and state context to determine opportunities for improvement.

Use of specialty OB consults during high-risk pregnancies in a Medicaid-covered population: initial impact of the Arkansas ANGELS intervention

This study examines the impact of a Medicaid-supported intervention (Antenatal and Neonatal Guidelines, Education and Learning System) to expand a high-risk obstetrics consulting service on the use of specialty consults between 2001 and 2006. Using a Medicaid claims-birth certificate data set, we find a decline over time in use of specialty consults for lower risk diagnoses and a shift to remote modalities for contact. Local physician participation in grand rounds via teleconference was associated both with specialty contact and use of remote modalities. Local physician use of a Call Center service was also associated with patient specialty contact. Expansion of telemedicine remote sites did not increase the likelihood of contact but was associated with the shift toward remote modalities. Specialty consult use and modality were influenced by the care context of the patient, particularly level of pregnancy risk, the specialty of the primary prenatal care provider, the timing of her prenatal care, and her ethnicity and education level.

Evaluating the effect of hospital and insurance type on the risk of 1-year mortality of very low birth weight infants: controlling for selection bias

OBJECTIVES

We examined the effect of hospital type and medical coverage on the risk of 1-year mortality of very low birth weight (VLBW) infants while adjusting for possible selection bias.

METHODS

The study population was limited to singleton live birth infants having birth weight between 500 and 1500 g with no congenital anomalies who were born in Arkansas hospitals between 2001 and 2007. Propensity score (PS) matching and PS covariate adjustment were used to mitigate selection bias. In addition, a conventional multivariable logistic regression model was used for comparison purposes.

RESULTS

Generally, all 3 analytical approaches provided consistent results in terms of the estimated relative risk, absolute risk reduction, and the number needed to treat. Using the PS matching method, VLBW infants delivered at a hospital with a neonatal intensive care unit (NICU) were associated with a 35% relative decrease (95% bootstrap confidence interval, 18.5%-48.9%) in the risk of 1-year mortality as compared with those infants delivered at non-NICU hospitals. Furthermore, our results showed that on average, 16 VLBW infants (95% bootstrap confidence interval, 11-32), would need to be delivered at a hospital with an NICU to prevent 1 additional death at 1 year. However, there was not a difference in the risk of 1-year mortality between VLBW infants born to Medicaid-insured versus non-Medicaid-insured women.

CONCLUSIONS

Estimated relative risk of infant mortality was significantly lower for births that occurred in hospitals with an NICU; therefore, greater efforts should be made to deliver VLBW neonates in an NICU hospital.

Quantifying risks of preterm birth in the Arkansas Medicaid population, 2001-2005

OBJECTIVE

The objective of this study was to examine risks of preterm births, quantify the explanatory power achieved by adding medical and obstetric risk factors to the models and to examine temporal changes in preterm birth due to changes in Medicaid eligibility and the establishment of a maternal-fetal medicine referral system.

STUDY DESIGN

The study used data from the 2001 to 2005-linked Arkansas (AR) Medicaid claims and birth certificates of preterm and term singleton deliveries (N=89 459). Logistic regression modeled the association among gestational age, demographic characteristics and risk factors, pooled and separately by year.

RESULT

Physiological risk factors were additive with demographic factors and explained more of the preterm birth ≤32 weeks than later preterm birth. Changing eligibility requirements for Medicaid recipients and increasing the financial threshold from 133 to 200% of federal poverty level had an impact on temporal changes. The proportion of births ≤32 weeks declined to 33%, from 3.0 to 2.0. However, later preterm births declined and then increased in the last year.

CONCLUSION

Physiological conditions are strongly associated with early preterm birth. Maternal behaviors and other stressors are predictive of later preterm birth. Unmeasured effects of poverty continue to have a role in preterm birth. Further examination of the referral system is needed.

Improving perinatal regionalization for preterm deliveries in a Medicaid covered population: initial impact of the Arkansas ANGELS intervention

OBJECTIVE

To examine the factors associated with delivery of preterm infants at neonatal intensive care unit (NICU) hospitals in Arkansas during the period 2001-2006, with a focus on the impact of a Medicaid supported intervention, Antenatal and Neonatal Guidelines, Education, and Learning System (ANGELS), that expanded the consulting capacity of the academic medical center's maternal fetal medicine practice.

DATA SOURCES

A dataset of linked Medicaid claims and birth certificates for the time period by clustering Medicaid claims by pregnancy episode. Pregnancy episodes were linked to residential county-level demographic and medical resource characteristics. Deliveries occurring before 35 weeks gestation (n=5,150) were used for analysis.

STUDY DESIGN

Logistic regression analysis was used to examine time trends and individual, county, and intervention characteristics associated with delivery at hospitals with NICU, and delivery at the academic medical center.

PRINCIPAL FINDINGS

Perceived risk, age, education, and prenatal care characteristics of women affected the likelihood of use of the NICU. The perceived availability of local expertise was associated with a lower likelihood that preterm infants would deliver at the NICU. ANGELS did not increase the overall use of NICU, but it did shift some deliveries to the academic setting.

CONCLUSION

Perinatal regionalization is the consequence of a complex set of provider and patient decisions, and it is difficult to alter with a voluntary program.

Medicaid and preterm births in Virginia: an analysis of recent outcomes

OBJECTIVES

The study objectives were to evaluate adequacy of prenatal care and risk for preterm birth among Medicaid clients in Virginia and to determine if payment method is associated with the risk of preterm birth.

METHODS

Birth certificate data for the Commonwealth of Virginia for 2007 and 2008 were linked with Medicaid claims data. Analysis was limited to singleton births. Three payment methods were evaluated: private insurance, self-pay, and Medicaid. The prevalence of preterm birth for each level of prenatal care defined by the Kotelchuck prenatal care index was assessed for each payment method. Unconditional logistic regression modeling was used to assess the association between payment method and preterm birth risk while controlling for known preterm birth risk factors.

RESULTS

Preterm birth prevalences (95% confidence interval [CI]) for the different payment methods were 7.9% (4.79-8.07) for the privately insured, 10.1% (9.57-10.60) for the self-pay group, and 10.2% (9.95-10.45) for Medicaid recipients. Compared with those with private insurance, women on Medicaid had an adjusted odds ratio (OR) for preterm birth (95% CI) of 0.99 (0.94-1.03). Self-pay mothers had a 32% increase in the odds of preterm birth relative to the privately insured. All payment groups show a trend toward significant reduction in preterm birth prevalence as adequacy of prenatal care improved from inadequate to adequate. Medicaid enrollees had a high prevalence of known risk factors, including smoking and illicit drug use and cervical insufficiency.

CONCLUSIONS

When known risk factors have been controlled, preterm birth risk for Medicaid enrollees did not differ significantly from the privately insured.

Late preterm infants: birth outcomes and health care utilization in the first year

OBJECTIVE

To distinguish the effects of late preterm birth from the complications associated with the causes of delivery timing, this study used propensity score-matching methods on a statewide database that contains information on both mothers and infants.

METHODS

Data for this study came from Arkansas Medicaid claims data linked to state birth certificate data for the years 2001 through 2005. We excluded all multiple births, infants with birth defects, and infants at <33 weeks of gestation. Late preterm infants (LPIs) (34 to 36 weeks of gestation) were matched with term infants (37-42 weeks of gestation) according to propensity scores, on the basis of infant, maternal, and clinical characteristics.

RESULTS

A total of 5188 LPIs were matched successfully with 15303 term infants. LPIs had increased odds of poor outcomes during their birth hospitalization, including a need for mechanical ventilation (adjusted odds ratio [aOR]: 1.31 [95% confidence interval [CI]: 1.01-1.68]), respiratory distress syndrome (aOR: 2.84 [95% CI: 2.33-3.45]), and hypoglycemia (aOR: 1.60 [95% CI: 1.26-2.03]). Outpatient and inpatient Medicaid expenditures in the first year were both modestly higher (outpatient, adjusted marginal effect: $108 [95% CI: $58-$158]; inpatient, $597 [95% CI: $528-$666]) for LPIs.

CONCLUSIONS

LPIs are at increased risk of poor health-related outcomes during their birth hospitalization and of increased health care utilization during their first year.