Prescription Opioid Use After Vaginal Delivery and Subsequent Persistent Opioid Use and Misuse

Abstract

Background:

Vaginal delivery is the most common reason for hospitalization in the United States and approximately 30% of women fill an opioid prescription after vaginal delivery, making this a common source of opioid exposure in women of reproductive age.

Objective(s):

To evaluate the effect of receiving an opioid prescription after vaginal delivery on the risk of subsequent persistent opioid use, opioid use disorders and overdose.

Study Design:

We assembled a nationwide cohort of Medicaid beneficiaries in the US using the Medicaid Analytic eXtract 2009 – 2014. The study population included pregnant women who delivered vaginally between 2009 and 2013 and were continuously enrolled in Medicaid from 90 days before to 365 days after delivery. We identified patients with prescription opioids dispensed within 7 days of the date of vaginal delivery. Persistent opioid use was defined as ≥ 10 opioid fills or >120 days’ supply dispensed from 30 to 365 days after delivery. Incident diagnoses of opioid use disorder and overdose were ascertained during the same interval. Propensity score matching was used to control for potential confounding.

Results:

Among 459,829 pregnancies ending in vaginal deliveries, 30.62% had an opioid dispensed within 7 days of delivery. Overall, 5,770 of the 140,807 (4.10%) women who filled an opioid prescription versus 2,668 of the 319,022 (0.84%) unexposed women had subsequent persistent opioid use, with an unadjusted relative risk (RR) of 4.90 (95% CI, 4.68 – 5.13) and a risk difference (RD) of 3.26% (3.15 – 3.37). After propensity-score matching, the risk remained higher among pregnancies with an opioid prescription dispensed, with a RR of 2.57 (2.43 – 2.72) and a RD of 2.21% (2.08 −2.33), which was confirmed by the instrumental variable analysis with a RD of 1.31% (1.06 – 1.56) by using the rate of opioid prescribing at the delivery facility in a given geographic region as the instrument. The adjusted RR of newly diagnosed opioid use disorder and overdose was 1.48 (1.40 – 1.57) and 1.92 (1.20 – 3.09) respectively.

Conclusion(s):

Opioid dispensing following vaginal delivery is associated with future persistent opioid use and misuse, independent of confounding factors. Opioid prescriptions to women after vaginal delivery should be avoided, except in rare circumstances.

INTRODUCTION

Vaginal delivery is the most common reason for hospitalization in the United States (US).¹ Approximately 30% of women fill an opioid prescription after vaginal delivery, making this a common source of exposure to opioids in young women.²

Prior studies suggest that opioid prescribing for acute indications may be associated with future chronic opioid use and even the development of opioid use disorder.^3–6 However, most of these studies have focused on clinical scenarios where pain is severe, for example post-surgical settings, and the use of prescription opioids may be justified to address acute pain that does not respond to other approaches. The US is the only country where opioids are routinely prescribed following vaginal delivery, while opioids are almost never used in other countries.⁷ Even within the US, there is significant heterogeneity in practice regarding opioid use following vaginal delivery. For example, opioid naïve women who deliver in the South are nearly five times more likely to be prescribed an opioid after vaginal delivery than women who deliver in the Northeast.²

The American College of Obstetricians and Gynecologists (ACOG) currently recommends nonsteroidal anti-inflammatory drug (NSAID) as a first-line agent for the treatment of pain following vaginal delivery and, if pain is inadequately treated with a NSAID alone, adding acetaminophen rather than an opioid.⁸ Given that opioid use for pain following vaginal delivery may be discretionary for many patients,^{9, 10} there is a particular need to define the risks associated with opioid exposure for this indication.

A few previous studies have reported that opioid use after vaginal delivery increases the risk for subsequent persistent opioid use. However, residual confounding by patients’ choice to fill an opioid prescription, by local opioid prescribing practices, and other factors may explain this association.^11–13 Moreover, none of the prior studies have evaluated the association between opioid exposure and the risk for new diagnoses of opioid use disorder or overdose, which are more clinically important outcomes than persistent use alone, and none focused on a nationwide sample of Medicaid patients, the population most at risk for opioid use disorder in the US.¹⁴ Therefore, we conducted a cohort study using the Medicaid Analytic eXtract (MAX) to examine the effect of prescription opioid use after vaginal delivery on the risk of subsequent persistent opioid use, opioid use disorder and overdose. We sought to overcome the potential limitations of prior studies by accounting for confounding by local opioid prescribing practices through adjustment for county level opioid prescribing rates and for confounding by patient preferences for opioid use by conducting a facility-based instrumental variable analysis.

MATERIALS AND METHODS

Data Source and Study Population

We assembled a cohort of pregnant women aged 12–55 years at the time of delivery using the nationwide MAX for the years 2009–2014. The MAX is comprised of a set of patient-level data files that provide information on patients’ Medicaid enrollment, demographics, outpatient and inpatient diagnoses and medical procedures, as well as medication dispensing records submitted for reimbursement. More details about the establishment of this pregnancy cohort have been described in previous literature.¹⁵

We included pregnant women who were continuously enrolled in Medicaid from 90 days before to 365 days after delivery. We excluded patients who had 1) opioid prescriptions with a cumulative days’ supply more than seven days in the 90 days before the date of delivery in order to exclude women who potentially used opioids chronically prior to delivery; 2) diagnosis or procedure codes for opioid use disorder, opioid poisoning, administration of opioid replacement therapy, or an outpatient prescription of naltrexone, naloxone or buprenorphine dispensed in the 90 days before the date of delivery; and 3) diagnosis or procedure codes for cesarean delivery in the 90 days before or 7 days after the date of delivery. For the purpose of controlling facility related confounding, we further required the deliveries to have a valid National Provider Identifier in MAX and excluded deliveries at a facility with less than 50 deliveries during the study period to have an accurate estimate of facility opioid prescribing preference for both the main and sensitivity analyses (eTable 1, and see Sensitivity Analysis for more details).

Exposure and Reference Groups

We determined exposure to opioids for vaginal delivery related pain based on the pharmacy dispensing records in MAX. A pregnant woman was considered exposed if she had an opioid prescription dispensed between the date of delivery and the first 7 days after delivery; women were classified as unexposed if they did not have a prescription opioid dispensed during this interval. Among the women who did not fill an opioid prescription after vaginal delivery, we further identified women who had a prescription dispensed for acetaminophen or NSAIDs during the first 7 days, which were considered as the active comparator group in a sensitivity analysis.

Outcomes

The primary outcome was persistent opioid use, defined as having ten or more opioid prescription fills on different dates or more than 120 days’ supply of opioids between 31 and 365 days after delivery.⁶ We included tablets and capsules only and allowed medication stockpiling, i.e. extended days’ supply for early refills. To test the robustness of our primary outcome, we varied the definition of persistent opioid use by requiring more than 180 days’ supply of opioid prescriptions between 31 and 365 days after delivery. Other secondary outcomes included the development of a new opioid use disorder based on having at least one inpatient or outpatient visit with relevant ICD-9 diagnosis codes, and opioid overdose based on having at least one inpatient or emergency department visit with relevant diagnosis codes.¹⁶

Covariates

We extracted a broad range of potential confounders and proxies for confounders from MAX. Covariates measured during the delivery hospitalization included maternal demographics (i.e., maternal age at delivery, race/ethnicity, region), complications related to vaginal delivery (i.e., trauma to perineum and vulva, high vaginal laceration, third-degree laceration/anal sphincter tear, fourth-degree laceration and episiotomy), and concurrent procedures (i.e., bilateral tubal ligation). Covariates measured during 90 days prior to the date of delivery included chronic comorbid conditions (i.e., depression, anxiety, post-traumatic stress disorder, sleep disorders), obstetric comorbidity score,¹⁷ other opioid indications (i.e., sciatica, arthritis/arthropathies, musculoskeletal pain, back/neck pain, migraine/headache, orthopedic injury, neuropathy or neuralgia, unclassified pain, fibromyalgia, and infection), non-opioid substance use disorder or dependence (i.e., alcohol abuse or dependence, tobacco use, cocaine use, marijuana use and amphetamine use), opioid use (i.e., number of opioid prescriptions with a cumulative days’ supply no more than seven days in the 90 days prior to delivery, total morphine equivalents for those opioid prescriptions in the 90 days prior to delivery), and other medications (i.e., antidepressants, benzodiazepines, other hypnotics, barbiturates, anxiolytics, mood stabilizers, typical antipsychotics, atypical antipsychotics and stimulants).

Aside from patient level confounding, there is concern for confounding introduced by high opioid prescribing settings; that is, patients who receive healthcare services in communities with high rates of opioid prescribing may be more likely to be prescribed opioids after vaginal delivery as well as be prescribed opioids during the follow-up period. To address this concern, we obtained the county-level Centers for Disease Control and Prevention (CDC) opioid prescribing rate for each facility by linkage based on the zip code.¹⁸ The county-level prescribing rate is defined as the total number of retail opioid prescriptions dispensed per 100 persons per year. For facilities with missing CDC county-level prescribing rates, which affected approximately 4% of the pregnancies in our cohort, we imputed the prescribing rate based on the state median.

Analyses

In unadjusted analyses, we calculated the absolute risks, risk differences and relative risks (RR) with 95% confidence intervals (CIs) to compare the effect of opioid exposure after vaginal delivery versus non-exposure on primary and secondary outcomes. In adjusted analyses, we estimated propensity scores (PS) for pregnancies with an opioid prescription dispensed after vaginal delivery versus unexposed pregnancies with logistic regression including all potential confounding variables listed above and performed 1:1 matching used a nearest-neighbor algorithm with matching caliper of 0.01 on the PS scale. We examined the covariate balance between groups before and after PS matching. An absolute standardized mean difference less than 0.1 was considered to indicate covariate balance. RRs were estimated using generalized linear regression models with log link and RDs were estimated using generalized linear regression models with identity link (PROC GENMOD in SAS [SAS Institute]). The research was approved by the institutional review board of Brigham and Women’s Hospital (IRB 2017P001660).

Sensitivity Analyses

To test the robustness of our findings, we conducted several sensitivity analyses. We first restricted the cohort to opioid naïve patients defined as not having filled an opioid prescription in the 90 days before the date of delivery, to rule out the possibility that subsequent persistent opioid use or misuse was due to previous opioid exposure. To address concerns about residual confounding, we identified an active comparator group consisting of patients who had at least one NSAID or acetaminophen prescription dispensed, and no opioid prescription dispensed between the date of delivery and the first 7 days after delivery. Lastly, to control for potentially unmeasured confounders, in particular patients’ preferences regarding opioid use (i.e., patients who chose to fill opioid prescriptions after vaginal delivery may be more likely to fill opioids for other indications during the follow-up) we performed a facility-based instrumental variable analysis (IVA) using a 2-stage least squares approach.¹⁹ To define the instrument, facilities were ranked within region according to their rate of opioid dispensed after vaginal delivery as measured during the study period and divided into deciles, which were used as the IV. To evaluate the strength of the instrument, we reported the crude proportion of pregnancies with opioid dispensings after vaginal delivery in each decile of the IV, the adjusted R² values for overall model fit, and the partial F statistics.¹⁹ More details on the IV analysis can be found in the Appendix eFigure 1.

RESULTS

There were 459,829 eligible women who underwent vaginal delivery, of whom 140,807 (30.6%) had an opioid dispensed after delivery (eTable 1). Compared with the unexposed, women receiving an opioid prescription after vaginal delivery were more likely to be White, smokers, reside in the South and in counties with higher opioid prescribing rates. They were also more likely to have had bilateral tubal ligation during the delivery hospitalization, depression, anxiety, arthritis, back or neck pain, migraine and other pain diagnosis, and use of opioids and other psychiatric medications in the 90 days before delivery. After PS matching, pre-specified covariates were well-balanced between groups, with a standardized difference < 0.1 for all covariates (Table 1).

Table 1.

Cohort characteristics of pregnancies with or without exposure to opioids after vaginal delivery in Medicaid Analytic eXtract 2009–2013

Characteristic	Before matching					After matching
	Opioid prescription after delivery				Standardized difference	Opioid prescription after delivery				Standardized difference
	Exposed N=140,807		Unexposed N=319,022			Exposed N=118,508		Unexposed N=118,508
	N	%*	N	%*		N	%*	N	%*
Year
2009	35992	25.6	78344	24.6	0.02	30249	25.5	30032	25.3	0.00
2010	33891	24.1	75188	23.6	0.01	28436	24.0	28587	24.1	0.00
2011	34793	24.7	82255	25.8	−0.03	29722	25.1	30203	25.5	−0.01
2012	19598	13.9	54061	16.9	−0.08	16339	13.8	15760	13.3	0.01
2013	16533	11.7	29174	9.1	0.09	13762	11.6	13926	11.8	0.00
Maternal demographic
Age, mean (SD)	24.35	5.5	24.72	5.7	−0.07	24.29	5.4	24.27	5.7	0.00
Race/Ethnicity †
White	77327	54.9	124874	39.1	0.32	60538	51.1	60767	51.3	0.00
Black	42153	29.9	92261	28.9	0.02	37594	31.7	38668	32.6	−0.02
Hispanic	14086	10.0	69073	21.7	−0.32	13477	11.4	12433	10.5	0.03
Other/Unknown	7241	5.1	32814	10.3	−0.19	6899	5.8	6640	5.6	0.01
Region
Midwest	44109	31.3	93421	29.3	0.04	40426	34.1	41871	35.3	−0.03
Northeastern	12800	9.1	108743	34.1	−0.64	12448	10.5	11002	9.3	0.04
South	63115	44.8	54477	17.1	0.63	46031	38.8	46317	39.1	−0.01
West	20783	14.8	62381	19.6	−0.13	19603	16.5	19318	16.3	0.01
CDC county level opioid prescribing rate, per 100 people, Mean (SD)	104.83	49.5	70.56	40.0	0.76	96.92	42.7	97.68	47.8	−0.02
Delivery characteristics and opioid indications
Bilateral tubal ligation	12325	8.8	4015	1.3	0.35	5103	4.3	3662	3.1	0.06
Periurethral trauma	5880	4.2	13022	4.1	0.01	5168	4.4	5477	4.6	−0.01
High vaginal laceration	1801	1.3	4582	1.4	−0.01	1597	1.4	1660	1.4	−0.01
Third-degree laceration or anal sphincter tear	1919	1.4	3663	1.1	0.02	1554	1.3	1586	1.3	0.00
Fourth-degree laceration	859	0.6	716	0.2	0.06	538	0.5	494	0.4	0.01
Episiotomy	16251	11.5	32793	10.3	0.04	13261	11.2	12646	10.7	0.02
Chronic comorbid conditions
Depression	6808	4.8	12024	3.8	0.05	5465	4.6	5671	4.8	−0.01
Anxiety	3754	2.7	5947	1.9	0.05	2930	2.5	2951	2.5	0.00
Post Traumatic Stress Disorder	657	0.5	1136	0.4	0.02	536	0.5	564	0.5	0.00
Sleep disorders	506	0.4	577	0.2	0.03	375	0.3	373	0.3	0.00
Obstetric comorbidity score	0.67	1.2	0.53	1.1	0.12	0.62	1.2	0.65	1.3	−0.02
Pain related conditions
Sciatica	605	0.4	1079	0.3	0.02	480	0.4	496	0.4	0.00
Arthritis/arthropathies, and musculoskeletal pain	6106	4.3	10239	3.2	0.06	4767	4.0	4918	4.2	−0. 01
Back or neck pain	11309	8.0	16984	5.3	0.11	8724	7.4	8872	7.5	−0.01
Migraine	5826	4.1	8881	2.8	0.07	4481	3.8	4582	3.9	0.00
Orthopedic injury	2840	2.0	4803	1.5	0.04	2310	2.0	2426	2.1	−0.01
Neuropathy /neuralgia	891	0.6	1436	0.5	0.03	683	0.6	736	0.6	−0.01
Fibromyalgia	560	0.4	767	0.2	0.03	437	0.4	434	0.4	0.00
Other pain	2451	1.7	1413	0.4	0.13	1407	1.2	1230	1.0	0.01
Infection	12542	8.9	28457	8.9	0.00	10613	9.0	10738	9.1	0.00
Substance use disorder or dependence
Alcohol abuse or dependence	386	0.3	820	0.3	0.00	324	0.3	359	0.3	−0.01
Tobacco use	23369	16.6	28723	9.0	0.23	17174	14.5	17508	14.8	−0.01
Cocaine use	156	0.1	284	0.1	0.01	125	0.1	138	0.1	0.00
Marijuana use	738	0.5	1517	0.5	0.01	604	0.5	645	0.5	−0.01
Amphetamine use	199	0.1	388	0.1	0.01	176	0.2	190	0.2	0.00
Medications
Number of opioid prescriptions in the 90 days prior to index
0	127072	90.2	303847	95.2	−0.19	108503	91.6	108411	91.5	0.00
1	11617	8.3	13747	4.3	0.16	8711	7.4	8899	7.5	−0.01
2	1857	1.3	1304	0.4	0.10	1162	0.9	1081	0.9	0.01
3+	261	0.2	124	0.0	0.04	132	0.1	117	0.1	0.00
Total morphine equivalents for all opioid prescriptions	32.1	1723.3	31.6	1848.1	0.00	30.1	1726.3	25.8	1364.7	0.00
Antidepressant	6907	4.9	7415	2.3	0.14	4900	4.1	4910	4.1	0.00
Benzodiazepine	1101	0.8	866	0.3	0.07	674	0.6	628	0.5	0.01
Other hypnotics	6161	4.4	7876	2.5	0.11	4546	3.8	4493	3.8	0.00
Barbiturates	1607	1.1	1408	0.4	0.08	1030	0.9	1016	0.9	0.00
Anxiolytics	425	0.3	307	0.1	0.05	250	0.2	238	0.2	0.00
Mood stabilizers	534	0.4	699	0.2	0.03	379	0.3	369	0.3	0.00
Typical antipsychotics	60	0.0	126	0.0	0.00	50	0.04	43	0.04	0.00
Atypical antipsychotics	468	0.3	669	0.2	0.02	364	0.3	365	0.3	0.00
Stimulants	439	0.3	388	0.1	0.04	307	0.3	293	0.3	0.00

^*Percentages may not total 100 because of rounding. Abbreviation: CDC - Centers for Disease Control and Prevention.

^†Race/ethnicity was determined based on information submitted to the Centers for Medicare & Medicaid Services by individual states, which was based on information that had been collected and coded from Medicaid applications. Other/Unknown category includes American Indian or Alaskan native, Asian, native Hawaiian or other pacific islander, more than one race, and unknown.

Risk of persistent opioid use and misuse

The risk of persistent opioid use was 4.10% (95% CI, 3.99 to 4.20) among pregnancies with an opioid prescription dispensed versus 0.84% (0.80 to 0.87) among the unexposed pregnancies, corresponding to an unadjusted RR of 4.90 (4.68 to 5.13) and a RD of 3.26% (3.15 −3.37). After matching, the RR attenuated to 2.57 (2.43 to 2.72) and the RD was 2.21% (2.08 to 2.33). For the secondary definition of persistent opioid use, i.e., requiring ≥180 days’ supply of opioids during follow-up, the risk was 1.83% (1.76 to 1.90) among pregnancies with an opioid dispensed after vaginal delivery and 0.35% (0.33 to 0.37) among the unexposed pregnancies. The corresponding unadjusted RR was 4.84 (4.59 to 5.09) versus the unexposed deliveries and the unadjusted RD was 1.48% (1.41 – 1.55). The adjusted RR attenuated to 2.54 (2.38 to 2.71), and the adjusted RD to 1.01% (0.93 to 1.10).

Among pregnancies with an opioid prescription dispensed after vaginal delivery, the risk of opioid use disorder was 2.78% (2.69 to 2.86) versus 1.20% (1.16 to 1.24) among the unexposed pregnancies. The unadjusted RR was 2.31 (2.21 to 2.41) and the RD was 1.57% (1.48 to 1.67). After PS matching, the adjusted RR was 1.48 (1.40 to 1.57) and the adjusted RD was 0.81% (0.69 to 0.92).

The risk of opioid overdose was 0.04% (0.03 to 0.05) among pregnancies with an opioid prescription dispensed after vaginal delivery and 0.02% (0.01 to 0.02) in the unexposed group, leading to an unadjusted RR of 2.19 (1.54 to 3.13) and a RD of 0.02% (0.01 to 0.04). After PS matching, the adjusted RR was 1.92 (1.20 to 3.09) and the adjusted RD was 0.02% (0.01 to 0.03) (Table 2 -3, Figure 1).

Table 2.

Absolute risks of persistent opioid use and misuse among mothers with or without exposure to opioids after vaginal delivery in Medicaid Analytic eXtract 2009–2013

Outcome	Opioid exposed No. of pregnancies =140,807		Unexposed No. of pregnancies =319,022
	Events, No. of Affected Pregnancies	Risk, No./100 Pregnancies (95% CI)	Events, No. of Affected Pregnancies	Risk, No./100 Pregnancies (95% CI)
Primary outcome
Persistent opioid use: 10 or more opioid prescription fills OR > 120 days’ supply of opioids	5,770	4.10 (3.99–4.20)	2,668	0.84 (0.80–0.87)
Secondary outcome
Persistent opioid use: ≥180 days’ supply of opioids	2,572	1.83 (1.76–1.90)	1,112	0.35 (0.33–0.37)
Opioid use disorder	3,908	2.78 (2.69–2.86)	3,833	1.20 (1.16–1.24)
Opioid overdose	60	0.04 (0.03–0.05)	62	0.02 (0.01–0.02)

Table 3.

Risk differences of persistent opioid use and misuse among mothers with or without exposure to opioids after vaginal delivery in Medicaid Analytic eXtract 2009–2013

Analysis	Risk Difference, % (95% CI)
Primary outcome
≥10 opioid prescription fills/ >120 days’ supply
Unadjusted	3.26 (3.15 – 3.37)
1:1 propensity score matching	2.21 (2.08 – 2.33)
Instrumental variable analysis	1.31 (1.06 – 1.56)
Secondary outcome
≥180 days’ supply opioids
Unadjusted	1.48 (1.41 – 1.55)
1:1 propensity score matching	1.01 (0.93 – 1.10)
Instrumental variable analysis	0.57 (0.39 – 0.74)
Opioid use disorder
Unadjusted	1.57 (1.48 – 1.67)
1:1 propensity score matching	0.81 (0.69 – 0.92)
Instrumental variable analysis	1.26 (1.02 – 1.50)
Overdose
Unadjusted	0.02 (0.01 – 0.04)
1:1 propensity score matching	0.02 (0.01 – 0.03)
Instrumental variable analysis	0.04 (0.01 – 0.07)

Figure 1.

Relative risks of persistent opioid use and misuse among mothers with or without exposure to opioids after vaginal delivery in Medicaid Analytic eXtract 2009–2013

Note: abbreviations, NSAIDs - Nonsteroidal anti-inflammatory drugs

Sensitivity analyses

After restricting the cohort to opioid naïve patients, the results were consistent with the main analysis. By using patients exposed to acetaminophen/NSAIDs as the active comparator, we also observed a similar magnitude of increased risks for both the primary and secondary outcomes (Figure 1).

There was marked variation between facilities in the frequency of opioid prescribing after vaginal delivery. The proportion of women who were dispensed an opioid after vaginal delivery in the bottom decile (low prescribing facility) was 4.6% versus 74.4% in the top decile (high prescribing facility) (eFigure 2). In the Stage 1 model of the IVA, compared to the model with baseline covariates only, adding the IV (i.e., deciles of facility prescribing preference) increased the adjusted R² from 0.19 to 0.32, and led to an improvement to the F statistic, indicating the IV was a strong predictor of opioid exposure after vaginal delivery. The IV analysis suggested an increase in the absolute risk for persistent opioid use associated with opioid exposure after vaginal delivery [RD for primary definition, 1.31% (1.06 to 1.56); secondary definition of persistent opioid use, 0.57% (0.39 to 0.74)], consistent with the primary PS matched analysis, although somewhat lower in magnitude. The increased risk for both opioid use disorder and overdose was also observed in the IV analysis, with a RD of 1.26% (1.02 to 1.50) and 0.04% (0.01 to 0.07) respectively (Table 3).

DISCUSSION

Principal Findings

In this nationwide cohort of Medicaid-insured pregnant women who underwent vaginal delivery, we found that approximately one third of them filled at least one opioid prescription after delivery. We observed a 2.5-fold increased risk for persistent opioid use associated with opioid exposure after vaginal delivery compared to non-exposure, and a near 2-fold increased risk for both opioid use disorder and opioid overdose. Given that there were approximately 2.5 million pregnancies that ended in vaginal deliveries in 2018 in the U.S. and among them 750,000 with an opioid prescription dispensed after vaginal delivery,²⁰ our findings indicate 16,575 additional cases of persistent opioid use, 6075 additional cases of opioid use disorder, and 150 additional cases of opioid overdose attributable to opioids that are dispensed after vaginal delivery.

Results

Our study confirms and extends findings from prior studies suggesting that opioid exposure after vaginal delivery is associated with persistent use. A prior study using Medicaid data from Tennessee suggested that opioid exposure after vaginal delivery was associated with an approximately 2.5-fold increase in the risk of persistent opioid use in the year of follow-up.¹¹ A recent study using the same Tennessee Medicaid data found that the risks of a composite of persistent opioid use, opioid use disorder diagnosis, buprenorphine or methadone prescription fill, opioid overdose diagnosis, or opioid-related death increased with increasing number of opioid prescriptions after vaginal delivery.²¹ Another study using nationwide commercial insurance data found that women dispensed an opioid near vaginal delivery (1 week before delivery to 3 days after discharge) had a significantly higher frequency of receiving at least one opioid prescription between 4 and 90 days and one or more prescriptions between 91 and 365 days after delivery.¹³ Extending these findings, we quantified the risks of individual outcomes and found that opioid exposure after vaginal delivery was associated not only with persistent opioid use, but also with an increased risk for the development of new opioid use disorder and opioid overdose. Our study also employed multiple advanced analytic approaches that together substantially increase the confidence that subsequent persistent use is attributable to receiving opioid prescriptions after vaginal delivery. We accounted for a wider range of confounding variables in our PS matched analysis than prior studies including, importantly, the intensity of local opioid prescribing as measured by county level opioid prescribing rates. Our study also extends prior work by evaluating a nationwide cohort of Medicaid beneficiaries. This is an important population to study as Medicaid covers approximately half of all pregnancies in the US.²² Further, Medicaid beneficiaries are at higher risk for the development of opioid use disorder. Data suggest that Medicaid is the primary payer for 75.0% of maternal stays related to substance use, whereas private insurance is the primary payer for 13.7% of the stays.²³

Clinical Implications

Despite the fact that opioids are not the first-line agent recommended by ACOG for the treatment of pain following vaginal delivery, we found that approximately 30% of the pregnant women filled at least one prescription for opioids after delivery, consistent with the frequency previously observed in a nationwide commercially insured population.² Importantly, our study documented that opioid prescribing practices are highly variable. Fifty-four percent of women who delivered vaginally in the South filled an opioid prescription, compared to 11% in the Northeast, 25% in the West, and 32% in the Midwest. Only 4.6% of women who delivered in a hospital in the lowest decile for prescribing in the IV analysis filled an opioid prescription compared to 74% in the top decile. Survey data suggest that the US is the only country in which opioids are regularly prescribed after vaginal delivery.⁷ Our finding that opioid exposure after vaginal delivery is linked to future persistent use, misuse, and overdose suggests an urgent need to reevaluate opioid prescribing practices following vaginal delivery, particularly in regions and hospitals with a high frequency of opioid prescribing. Further, while not measurable in our data, it is known that leftover medication is a major source for prescription opioids that are diverted or misused.²⁴ Additionally, overdose is one of the emerging contributors to maternal death,²⁵ and opioids are the most common drug identified in the toxicology testing of women with deaths from self-harm.²⁶ Given the large number of opioid prescriptions filled after vaginal deliveries in the US, excessive opioid prescribing is therefore likely an important contributor to leftover opioids being introduced into communities and leading to opioid overdose and maternal mortality.

Research Implications

Due to the long lag time in the release of nationwide Medicaid data from the Centers for Medicare and Medicaid Services, we were unable to examine the opioid prescribing patterns after year 2014. Initiatives have been developed to decrease postpartum opioid use in some areas and hospitals in more recent years.^27–29 Our findings highlighted the importance of such changes in high opioids prescribing practice after routine vaginal delivery, and future research should be focused on evaluating the effectiveness of those quality improvement programs to reduce the unnecessary opioid use after vaginal delivery.

Strengths and Limitations

Our study has a number of important strengths including the large size of our database and the nationwide scope of the MAX sample. Opioid exposure was defined based on pharmacy dispensing records which avoids recall bias. We used multiple approaches to control for factors that could confound the causal link between opioid exposure after vaginal delivery and persistent use including PS based methods to adjust for measured confounders, an active comparator group analysis, and an IV analysis to address potential concerns for unmeasured confounding.

The study is also subject to certain limitations inherent in its design. First, while we are able to observe filled prescriptions, we do not know if women actually used the medication. The focus of our study, however, was to evaluate the effect of prescribing/dispensing opioid after vaginal delivery, regardless of whether the opioids were actually used by the patients. In addition, since we required fixed follow-up to measure persistent opioid use, we did not include patients who died before the end of follow-up, including fatal overdose, which rarely occurs. Another potential limitation is the possibility of residual confounding by unmeasured variables that account for the increased risk. In this regard, we were particularly concerned about potential unmeasured confounding due to patient preference and attitudes towards opioid use or the level of pain experienced, i.e. patients who tend to fill an opioid prescription after vaginal delivery may also be more likely to use opioids persistently or misuse opioids in the long term. Given the nature of claims database, we were unable to determine the reason for long term use. To address concerns about residual confounding, we conducted an IV analysis by using the facility opioid prescribing preference as the IV. The results of this analysis were generally consistent with those of the main analysis, and therefore strengthened the confidence in our findings that opioid dispensing after vaginal delivery leads to an increased risk for persistent opioid use, opioid use disorders and overdose. On the other hand, facility preference based IV may be confounded by case mix and other facility related factors. To alleviate this concern, we adjusted for potential confounders, including patient characteristics, in the IV analyses.

Conclusions

Among publicly insured pregnant women, a large fraction filled opioid prescriptions after vaginal delivery, which is associated with future persistent opioid use and misuse, independent of confounding factors. This is important from a public health perspective, as vaginal delivery is the most common reason for hospitalization in the US. Given this risk, prescription opioid use after vaginal deliveries should not be routinely prescribed and non-opioid analgesics should be used preferentially for the treatment of pain following vaginal delivery, except in rare circumstances.

1. Why was the study conducted?

   While approximately 30% of women fill an opioid prescription after vaginal delivery in the United States, the effect of receiving an opioid prescription after vaginal delivery on the risk of subsequent persistent opioid use, opioid use disorders and overdose is unknown.

2. What are the key findings?

   Receiving an opioid prescription after vaginal delivery was associated with a 2.5-fold increased risk for persistent opioid use, and a near 2-fold increased risks for opioid use disorder and opioid overdose.

3. What does this study add to what is already known?

   Using advanced methods to reduce residual and unmeasured confounding, we conclude that opioid prescriptions to women after vaginal delivery should be avoided, except in rare circumstances.