Is it good to be too light? Birth weight thresholds in hospital reimbursement systems☆
Introduction
Small changes in birth weight can have important financial implications for hospitals in many of the widespread prospective payment schemes (PPS) which reimburse hospitals with a fixed rate for the treatment of strictly defined diagnosis related groups (DRGs). More specifically, hospitals receive a higher reimbursement for newborns with birth weight just below certain thresholds than for newborns with weight above, leading to an incentive to report lower birth weight. The evidence is accruing that the introduction of birth weight thresholds has led to large misreporting – so-called upcoding – of birth weight (e.g. Jürges and Köberlein, 2015, Shigeoka and Fushimi, 2014).
At the same time, birth weight thresholds are used to diagnose newborns as having “extremely low” (weight ≤1000 g), “very low” (weight ≤1500 g), or “low” birth weight (weight ≤2500 g) and appear in medical guidelines. Despite the fact that low birth weight is typically linked to worse health outcomes (see e.g. Hack et al., 2002, Hummer et al., 2014), newborns with weight just below the 1500 g diagnostic threshold have been found to have higher survival chances than newborns with weight just above (Almond et al., 2010, Bharadwaj et al., 2013, Breining et al., 2015). To the extent that reimbursement-relevant birth weight thresholds are identical to the diagnostic thresholds, the practice of upcoding newborns’ weight may thus be to the benefit of the newborn. Furthermore, as the hospital receives more money for newborns with weight below the reimbursement-relevant thresholds, they may also be able or willing to deliver additional care to these newborns.
In this paper, we investigate whether newborns benefit from having a reported birth weight below reimbursement-relevant weight thresholds in the German DRG system. Based on an administrative hospital claims data set covering the universe of hospital births in Germany in the years 2005–2011, we compare the survival chances and the quantity of care that newborns with weight just below the relevant thresholds receive to those of newborns with weight just above the thresholds. We include all eight reimbursement-relevant thresholds in the German DRG system. While some of these thresholds overlap with the diagnostic thresholds and/or are explicitly mentioned in medical guidelines, others are only relevant for reimbursement, allowing us to shed light on the importance of diagnostic thresholds, medical guidelines, and reimbursement for the care of newborns.
Different from the settings in the earlier literature on the effects of diagnostic thresholds (Almond et al., 2010, Bharadwaj et al., 2013, Breining et al., 2015), the fact that the diagnostic thresholds are relevant for reimbursement in our setting imposes a challenge to the empirical analysis: If the decision to upcode a newborn's weight below a threshold depends on the newborn's health status, the crucial assumption that newborns with reported weight above and below the threshold have ex ante similar health is not plausible, an issue that Barreca et al. (2011) already raised in light of the earlier literature. As Jürges and Köberlein (2015) show, it is likely that birth weight manipulation in German hospitals is not random. On the contrary, hospitals primarily tend to upcode newborns for whom staff expects higher care needs. These are relatively fragile newborns that still have non-negligible survival chances and will therefore receive a lot of treatment. We take four steps in our analysis to take this into account: First, we control for a large set of variables capturing a newborn's health at birth. We specifically choose variables such as sex and plurality of births that are not easy to manipulate and observable to the hospital staff who reports newborns’ weight. These are variables that may influence reported birth weight. Second, we restrict our analysis to newborns who survive the first day – and as a robustness check the first four days – of their life, thus excluding those very fragile cases for whom hospital staff may expect an early death, making upcoding of birth weight not worthwhile. In a third set of results, we further include hospital fixed effects in our estimations, taking into account possible differences in coding practices and treatment across hospitals. Fourth, as adding control variables and dropping newborns who likely have low survival chances at birth may not fully control selective upcoding, we apply an approach suggested by Gerard et al. (2016) to bound treatment effects in regression discontinuity settings when the running variable is manipulated.
Our paper contributes to and brings together two strands of literature. The first focuses on the question whether hospitals upcode diagnoses or other health measures or patient characteristics, such as birth weight, to generate higher payments in DRG reimbursement systems. Concerning the upcoding of diagnoses, Dafny (2005) shows that hospitals reacted to a recalibration of Medicare DRGs in 1988 by disproportionally shifting patients to diagnoses codes that became more lucrative. At the same time, she finds no changes in the treatment that patients receive nor in patient mortality. Silverman and Skinner (2004) focus on Medicare patients with respiratory disease and show that the share of patients coded to the highest paying DRG increases significantly over time, particularly so in the group of for-profit hospitals. Furthermore, there is evidence that hospitals in the Italian region of Emilia-Romagna (Verzulli et al., 2016), Portugal (Barros and Braun, 2016) and Norway (Januleviciute et al., 2016) upcode patients to the highest paying DRGs. Concerning upcoding of patient characteristics, Shigeoka and Fushimi (2014) show that hospitals in Japan have manipulated birth weight as a response to the introduction of a partial PPS in a way that increased hospital payments. Similarly, Jürges and Köberlein (2015) find that German hospitals reacted to the introduction of the DRG payment system in 2003 by systematically misreporting newborns’ weight. While Shigeoka and Fushimi (2014) investigate differences in treatment, such as length of stay or number of procedures that newborns receive for Japan, there is no evidence to date on the effects of birth weight manipulation on the quantity or quality of care that newborns receive in the German setting. These effects, however, are particularly important as differences in reimbursement around birth weight thresholds in Germany could add to treatment differences around birth weight thresholds resulting from medical guidelines or diagnostic thresholds, the impact of which is the focus of the second strand of literature.
This second strand of literature focuses on the effect of birth weight thresholds on the quantity and quality of care that newborns receive. Based on the census of U.S. births, Almond et al., 2010, Almond et al., 2011 find that newborns with weight just below the very low birth weight threshold at 1500 g have higher survival chances than newborns with weight just above. Based on hospital discharge data for five states they further find that birth weight below 1500 g triggers additional care that also results in higher hospital charges. Their results are particularly concentrated among low quality hospitals, i.e., those hospitals that offer no or only low levels of neonatal intensive care. Similar effects of the very low birth weight thresholds have been found for newborns in Chile where medical guidelines explicitly recommend different treatment depending on a very low birth weight diagnosis (Bharadwaj et al., 2013). Breining et al. (2015) focus on Denmark where the treatment recommendations in medical guidelines only vary across the very low birth weight thresholds for newborns with at least 32 weeks of gestation. They find that for newborns born at 32 weeks of gestation or more, treatment depends on birth weight. For newborns born earlier in the pregnancy they find no treatment differences with birth weight, indicating that medical guidelines have an impact on the care that newborns receive.
Our paper brings these two strands of the literature together in investigating whether newborns benefit from upcoding of birth weight below thresholds that may themselves affect the care that newborns receive as they are diagnostic thresholds. Our analysis contributes to the first strand of the literature by investigating effects of upcoding on the quantity and quality of care that patients receive. Focusing on the care received as a consequence of manipulation allows for conclusions concerning physician behavior, namely whether they are willing to misreport in order to prevent reimbursement rules from interfering with their treatment decisions. We add to the second strand of literature by focusing not only on the very low birth weight threshold but also on other thresholds that impact diagnoses and appear in medical guidelines.
Our findings show that without controls and not restricting to first day survivors newborns with weight below almost all of the eight thresholds stay in hospital longer, receive more procedures, and have lower mortality (during the hospital stay). However, for all but the highest two thresholds (2000 g and 2500 g) these results become insignificant or even change sign, when we control for health at birth, exclude newborns who die on the first day – or first four days – of life, and include hospital fixed effects. We interpret this as evidence that neither reimbursement differences, nor the diagnostic threshold of 1500 g or thresholds in medical guidelines trigger additional care or reduce mortality among newborns in Germany. These results are also supported by the bounds approach based on Gerard et al. (2016) with bounds that always include zero for all but the higher weight thresholds (2000 g and 2500 g) where this estimation is not possible. The significant differences in care that we find around these higher thresholds may thus reflect actual differences in care. For these thresholds we also investigate specific differences in the actual procedures with the finding that newborns with weight below the 2000 g and 2500 g threshold are more likely to be monitored for hearing problems and breathing as well as cardio-vascular problems and are also more likely to receive additional phototherapy and electrolyte solution application. We conclude that – at the lower weight thresholds and thus for the particularly frail cases – the care that a newborn receives does not seem to depend on the reimbursement that the hospital receives for their care.
These results provide insights into the economic behavior of physicians and other (medical) hospital staff. As further discussed in Section 2 birth weight in German hospitals is measured and reported by medical hospital staff – usually midwives, nurses or physicians – who also decide on the care that newborns receive. The evidence of birth weight manipulation implies that medical staff is aware of the birth weight thresholds. Nonetheless, our results show that these thresholds do not affect the care that newborns receive, at least not for high risk cases. This suggests that medical personel may use the option to manipulate birth weight in order to prevent reimbursement differences from interfering with medical decision making and thus to allow them to put their patients’ health and wellbeing first. Medical staff thus seems to be willing to misreport to allow them to act in line with their professional norms. In the final section of this paper we further discuss policy implications of these findings.
The rest of the paper is structured as follows: In Section 2, we give an overview on the institutional background in Germany and discuss which birth weight thresholds may be related to the quantity and quality of care that newborns receive. We introduce our data in Section 3 and the empirical strategy in Section 4. In Section 5, we present our results, the sensitivity of which we explore in Section 6. The paper closes with a discussion and conclusion in Section 7.
Section snippets
Institutional background
In this section we give a brief overview on the German DRG (G-DRG) system and then describe reasons why the treatment that newborns receive in German hospitals may vary around birth weight thresholds.
Data
Our analyses are based on the universe of German hospital claims from the years 2005–2011. All German hospitals have to submit their DRG claims to the Institute for Hospital Reimbursement (InEK). InEK forwards parts of the data to the German Federal Statistical Office, which makes the data available to researchers.11
For each of the roughly 20 million hospital stays per year in Germany, the data contain
Empirical strategy
We aim at investigating whether newborns benefit from having a (reported) birth weight below birth weight thresholds defined by the DRG system, medical guidelines or a combination thereof. We therefore document differences in quantity and quality of care comparing newborns with weight just below and just above the eight reimbursement relevant thresholds in Germany. To this end we start by estimating mean differences in 25 g weight intervals above and below the thresholds. Quantity of care is
Results
Table 2 reports differences in means between groups of 25 g intervals across the eight reimbursement-relevant birth weight thresholds (600 g, 750 g, 875 g, 1000 g, 1250 g, 1500 g, 2000 g, 2500 g), as well as for four thresholds that play no role in reimbursement or medical guidelines (700 g, 1300 g, 2200 g, 2700 g).19
Sensitivity analyses
In this section we present results of four sets of sensitivity analyses. First, we explore the sensitivity of the results to a so-called donut-regression and thus to excluding from our data newborns with weight very close to the thresholds. We conduct this analysis to investigate whether our results are sensitive to rounding to the thresholds. Second, we explore whether the choice of the polynomial in birth weight affects the results, and third whether the choice of bandwidth has an impact. As
Discussion and conclusion
In this paper, we investigate whether birth weight thresholds in hospital reimbursement systems affect the quantity and quality of care delivered to newborns. Using the universe of hospital births in Germany from the years 2005–2011, we document that newborns with weight below all but one reimbursement relevant threshold receive more care and have lower risk of dying during the hospital stay than their neighbors with weight at or above the respective thresholds. For all but the highest weight
References (35)
- et al.
What explains DRG upcoding in neonatology? The roles of financial incentives and infant health
J. Health Econ.
(2015) Manipulation of the running variable in the regression discontinuity design: a density test
J. Econom.
(2008)Health care reform and Diagnosis Related Groups in Germany: the mediating role of Hospital Liaison Committees for Jehovah's Witnesses
Soc. Sci. Med.
(2016)- et al.
Supplier-induced demand for newborn treatment: evidence from Japan
J. Health Econ.
(2014) - et al.
Medicare upcoding and hospital ownership
J. Health Econ.
(2004) - et al.
Estimating marginal returns to medical care: evidence from at-risk newborns
Q. J. Econ.
(2010) - et al.
The role of hospital heterogeneity in measuring marginal returns to medical care: a reply to Barreca, Guldi, Lindo, and Waddell
Q. J. Econ.
(2011) Folgen der Ökonomisierung im Krankenhaus für Ärzte täglich spürbar
(2017)S2k-Leitline der Gesellschaft für Neonatologie und pädiatrische Intensivmedizin: Frühgeborene an der Grenze der Lebensfähigkeit, AWMF Leitlinien-Register Nr. 024-019
(2007)S1-Leitline der Gesellschaft für Neuropädiatrie: VisuelleWahrnehmungsstörungen, AWMF-Leitlinien-Register Nr. 022-020
(2010)
S2k-Leitlinie der Gesellschaft für Neonatologie und pädiatrische Intensivmedizin: Nekrotisierende Enterokolitis, AWMF-Leitlinien-Register Nr. 024-009
S2k-Leitlinie der Gesellschaft für Neonatologie und pädiatrische Intensivmedizin: Betreuung von gesunden reifen Neugeborenen in der Geburtsklinik, AWMFLeitlinien-Register Nr. 024-005
S2k-Leitlinine der Deutschen Gesellschaft für Kinderchirurgie: Kurzstreckige Ösophagusatresie, AWMF Leitlinien-Register Nr. 045-006
S2k-Leitlinie der Deutschen Gesellschaft für Phoniatrie und Pädaudiologie: Periphere Hörstörungen im Kindesalter, AWMF-Leitlinien-Register Nr. 049-010
S3-Leitlinie der Deutschen Gesellschaft für Ernährungsmedizin: Parenterale Ernährung in der Kinder- und Jugendmedizin, AWMF Leitlinien-Register Nr. 073-023
Saving babies? Revisiting the effect of very low birth weight classification
Q. J. Econ.
Upcoding in a national health service: the evidence from Portugal
Health Econ.
Cited by (18)
Price and income effects of hospital reimbursements
2022, Journal of Health EconomicsCream skimming by health care providers and inequality in health care access: Evidence from a randomized field experiment
2021, Journal of Economic Behavior and OrganizationCitation Excerpt :For Germany, Schmitz (2013) shows that newly introduced practice-level budgets for the publicly insured reduced the likelihood of follow-up visits. Reif, Wichert and Wuppermann (2018) use administrative data on newborns and several discontinuities in birth weight thresholds that trigger higher reimbursement rates to test for whether treatments quantity and quality differs. While they find evidence for weight manipulation, they conclude that there is not much evidence for causal revenue-related changes in treatment quantity and quality at these margins in Germany.
Do hospitals respond to changing incentive structures? Evidence from Medicare's 2007 DRG restructuring
2020, Journal of Health EconomicsCitation Excerpt :They report an increase in manipulation of infant birthweights and also find longer stays in the neonatal intensive care unit, both of which are associated with greater reimbursement. Reif et al. (2018) examine if this upcoding of neonates is possibly good for the neonates in that it may lead to better care. However, they find no evidence that this is the case.
Birth weight, neonatal care, and infant mortality: Evidence from macrosomic babies
2020, Economics and Human BiologyFinancial Influence on Medical Decisions in Germany - A Review of the Reasons and State of Research on Economization in Inpatient Care
2023, Deutsche Medizinische Wochenschrift
- ☆
We are grateful for helpful comments and suggestions by Tom Crossley, Michael Gerfin, Hendrik Jürges, Harald Tauchmann, Marcos Vera-Hernandez, and Joachim Winter. Participants of the 2nd German Health Econometrics Group Workshop in Essen, the Workshop on Early Care Interventions and their Effects on Children and Families in Aarhus, the SSPH Doctoral Workshop in Health Economics and Policy in Lucerne, the 7th Annual Conference of the German Health Economics Association in Bielefeld, the 11th European Conference on Health Economics in Hamburg and the ISER seminar in Colchester provided valuable feedback. We also want to thank Melanie Scheller for assistance with data access and Leonard Goff for assistance with the rdbounds package. Funding through the International Doctoral Program “Evidence-Based Economics” of the Elite Network of Bavaria is gratefully acknowledged. Christina Schramm, Katrin Poschen and Katrin Ziegler provided valuable research assistance.