Indication-Specific Opioid Prescribing for US Patients With Medicaid or Private Insurance, 2017

Key Points Question What are indication-specific opioid prescribing rates for nonsurgical acute or chronic pain or postoperative pain conditions and pain related to cancer or sickle cell disease? Findings In this cross-sectional analysis examining data from 18 016 259 patients with private insurance and 11 453 392 patients with Medicaid benefits, opioid prescribing rates, days’ supply, and daily dosage varied widely across clinical indications and by insurance type. Meaning Potential inconsistencies between indication-specific prescribing patterns and relevant clinical recommendations highlight opportunities for selection of safer and more effective pain treatment options.

For both datasets, the overall study sample included all beneficiaries enrolled with both medical and prescription drug coverage for at least ten continuous months from July 1, 2016-March 31, 2018 (OLDW) and April 1, 2016-December 31, 2017 (MMD). These differing windows reflect the availability of only 2016 and 2017 MMD data for this analysis. Ten months of continuous enrollment comprised the month of the index diagnosis, including at least six months prior and at least three months after it (to determine whether the diagnosis was acute or persisted as chronic). Thus, the continuous enrollment period varied by individual enrollee depending on the date of the index diagnosis.

Indication inclusion criteria
Indications were chosen for analysis based on a combination of their association with opioid prescribing reported in the literature; feedback from the Opioid Prescribing Estimates Workgroup via the NCIPC BSC regarding clinical importance; and frequency in the OLDW dataset (i.e., rare conditions and procedures were excluded). Claims that contained the ICD-10-CM code for "Z51.5 encounter for palliative care" were excluded from analysis.

Indications associated with nonsurgical acute pain
We included abdominal pain, acute low back pain, acute migraine, dental pain, rib fractures, herpes zoster, renal colic, and musculoskeletal sprains/strains (ICD-10-CM codes are listed in eAppendix 2). For each acute pain indication, the diagnosis code itself is insufficient to determine whether the indication is of an acute or chronic nature. Thus, we first used ICD-10-CM diagnosis codes in professional claims from January 1, 2017-December 31, 2017 (OLDW) or October 1, 2016-September 30, 2017 (MMD) to identify all visits with a particular nonsurgical acute pain indication. Second, we excluded all identified visits that represented inpatient stays. Third, we defined the first diagnosis of a specific nonsurgical acute pain indication in 2017 (or starting with October 1, 2016, for MMD data) as the "index diagnosis" for a patient. Fourth, we used six months of claims prior to the index diagnosis to determine whether it was a new or pre-existing diagnosis. Fifth, we used claims three months after the "index diagnosis" to determine whether the index diagnosis persisted beyond this timeframe. Sixth, we only included visits that met the following criteria: 1) a new diagnosis, i.e. no same diagnosis in the six months prior, as determined during the fourth step; 2) a diagnosis that persisted three months or less, as determined during the fifth step; 3) no more than three visits for the same diagnosis within the three months following the "index diagnosis" during the study period. There was one exception in approach: in order to broadly capture abdominal pain complaints, we used the nonspecific abdominal pain symptom ICD-10-CM codes (R100, R101, R103, or R109) on a claim record to identify these visits.

Indications associated with chronic pain
We included non-radicular back pain, radicular back pain, neck pain, fibromyalgia, inflammatory joint disorders, irritable bowel syndrome, non-migraine headaches, osteoarthritis and joint cartilage conditions, and periarticular/soft tissue disorders (ICD-10-CM codes are in eAppendix 2). Similar to nonsurgical acute pain indications, the diagnosis code itself is insufficient to determine whether the indication is of a chronic nature. To identify chronic pain indications for our analysis, we first applied a similar process to that described earlier for nonsurgical acute pain. Thus, we first used ICD-10-CM diagnosis codes in professional claims from January 1, 2017-December 31, 2017 (OLDW) or October 1, 2016-September 30, 2017 (MMD) to identify all visits with a particular chronic pain indication. Second, we excluded all identified visits that represented inpatient stays. Third, we defined the first diagnosis of a specific chronic pain indication in 2017 (or starting with October 1, 2016, for MMD data) as the "index diagnosis" for a patient. Fourth, we used six months of claims prior to the index diagnosis to determine whether it was a new or pre-existing diagnosis. Fifth, we used claims three months after the "index diagnosis" to determine whether the index diagnosis persisted beyond this timeframe. However, different from the approach for nonsurgical acute pain, we then only included in our analysis those visits for which the chronic pain indication persisted beyond three months after the index diagnosis, as determined in the fifth step 1,2 .

Indications associated with postsurgical pain
We included 24 surgical procedures in our study (the list of procedures and their CPT codes are in eAppendix 2). Surgical procedures were identified by using the CPT code on professional claims, i.e. from providers who performed procedures. We included professional claims from January 1, 2017-December 31, 2017 (OLDW) or October 1, 2016-September 30, 2017 (MMD) from both outpatient and inpatient settings in order to capture procedures occurring in both settings.

SCD pain
SCD pain was analyzed separately from the other pain categories due to the inability to reliably differentiate between acute pain related to a sickle cell vaso-occlusive crisis and chronic SCD-related pain using claims data and because of the unique clinical characteristics and approach to pain management for this patient population. To analyze opioid prescriptions for pain related to SCD, we first applied the same steps as listed above for chronic pain. Then, as per prior studies 3,4 we only included those patients who had at least three distinct visits with an ICD-10-CM code for SCD during the study period (see eAppendix 2 for a list of ICD-10-CM codes).

Cancer pain
We analyzed opioid prescriptions written for cancer pain separately from the other categories above, due to the unique clinical characteristics and approach to pain management for this patient population. Cancer patients were identified as having at least two visits in January 1, 2017-December 31, 2017 (OLDW) or October 1, 2016-September 30, 2017 (MMD) with both a cancer diagnosis and a provider specialty listed as "oncologist" or "oncology." ICD-10-CM codes for this indication are listed in eAppendix 2.

Indications associated with postsurgical pain
For surgical procedures, we linked an opioid prescription to a procedure if the prescription met all of the following criteria: 1) the prescription and procedure had the same patient ID; 2) the prescription was dispensed within 5 days after the procedure including the day of procedure; and 3) there were no additional procedures or visits between the procedure in question and the date the prescription was issued.
Given the possibility of patients obtaining opioid prescriptions from surgeons prior to a scheduled procedure, we also applied additional criteria to link prescriptions to procedures: 1) the prescription and procedure had the same patient ID; 2) the prescription was dispensed within 30 days prior to the procedure; 3) the prescriber specialty was "surgeon;" and 4) the prescription was not linked to any other procedure or visit prior to the date of this prescription.

Indications associated with nonsurgical acute pain
For nonsurgical acute pain indications, we linked an opioid prescription to a visit if the prescription met all of the following criteria: 1) the prescription was not linked to any surgical procedure; 2) the prescription and visit had the same patient ID; 3) prescription was dispensed within 7 days after the visit, including the day of visit; and 4) there were no additional visits with the diagnosis in question between this visit and the date the prescription was issued.
To further ensure the accurate linkage of opioid prescriptions to nonsurgical acute pain indications, we searched for a patient's opioid prescriptions, if any, in the three months prior to his or her visit where the index diagnosis was made. A patient receiving long-term opioid therapy (LTOT) was defined as: 1) having at least three opioid prescriptions in the three months prior to the index visit; 2) more than 60 total days of opioid supply; and 3) the gap between the end of one prescription and the next prescription was fewer than 10 days. If a patient was identified as receiving LTOT prior to the index acute pain diagnosis, we excluded the patient from the nonsurgical acute pain category to increase confidence that a given opioid prescription was for the nonsurgical acute indication, since a patient with a new nonsurgical acute condition may present to the same clinician that prescribes LTOT for a chronic indication.
We conducted our analysis for nonsurgical acute pain at the visit level because of the nature of acute pain conditions, for which each presentation for care may represent distinct events. The majority of patients in this category had a single visit (i.e. the index diagnosis visit) for an acute pain condition between the date of index diagnosis and the three months following the index visit. However, in cases when a patient had more than one visit of the same nonsurgical acute pain condition, each visit was counted separately. When a prescription was linked to more than one nonsurgical acute pain indication, we first applied the provider ID and provider specialty, where it was available, to further assign the prescription. When it was not available, we assigned the prescription to the indication whose visit date was closer to the prescription.

Indications associated with chronic pain or SCD pain
For chronic pain indications and SCD pain, we linked an opioid prescription to a visit if the prescription met all of the following criteria: 1) the prescription was not linked to any surgical procedure or nonsurgical acute pain; 2) the prescription and visit had the same patient ID; 3) the prescription was dispensed within 14 days after the visit including the day of visit; and 4) there were no additional visits with the diagnosis in question between this visit and the date the prescription was issued. The wider window of 14 days was used because the filling of opioid prescriptions for chronic pain indications might not occur as rapidly as for acute pain, possibly because a patient might already have opioids on hand. In cases where a patient had two concurrent chronic or SCD pain indications that overlapped within the same time frame, the opioid prescription was assigned based on the likelihood tier to which the indication was assigned (see below). These likelihood tiers were constructed based on multiple published clinical guidelines [5][6][7][8][9][10] and, based on that published guidance, the likelihood that an opioid would be prescribed for that condition if guideline recommendations were followed. If two indications fell in the same tier, the prescription was assigned to whichever indication was always coded as the primary diagnosis. If primary indication could not be identified, the prescription was assigned to each indication. A = Opioids may be prescribed. Opioids may be considered for pain management if benefits are felt to outweigh risks, based on existing clinical guidelines. B = Opioids not likely to be prescribed. Based on existing clinical guidelines, opioids are not typically used for pain management as the risks are usually felt to outweigh the benefits.
Criteria to identify patients receiving LTOT are described above under "Indications associated with nonsurgical acute pain." A patient not receiving LTOT in our analysis was any patient that did not meet the criteria for being on LTOT.

Cancer pain
For cancer-related pain, opioid prescriptions were linked following a similar methodology to that described above for chronic pain indications and SCD-related pain. However, the prescription linkage for this category occurred as the primary step in the linkage algorithm, meaning that opioid prescriptions linked to management of cancer pain were thus excluded from the linkage process for other pain indications.

Opioid prescription linkage algorithm
Identify patients with cancer between 01/01/2017 and 12/31/2017 using the following criteria: 1) At least two visits during the study period with an ICD-10-CM code for a cancer diagnosis under study; 2) Provider specialty listed as "oncologist" or "oncology" 1) For each patient with cancer, identify the first claim of a cancer diagnosis during the study period, i.e. "index diagnosis." 2) Apply the inclusion criterion that the diagnosis must last at least 90 days after the index diagnosis

NO OPIOIDS in 3 months prior to date of procedure ("not on LTOT")
Link nonexcluded opioid prescription to the procedure, as follows: 1) Prescription filled within 5 days of the procedure, including the procedure day; 2) No other procedure between the linked procedure and the date of prescription filling; 3) Prescription filled within 30 days prior to the procedure AND prescribed by surgeon, with no other procedure prior to this prescription Apply 10-month continuous enrollment criteria Link nonexcluded opioid prescription to the procedure, as follows: 1) Prescription filled within 5 days of the procedure, including the procedure day; 2) No other procedure between the linked procedure and the date of prescription filling; 3) Prescription filled within 30 days prior to the procedure AND prescribed by surgeon, with no other procedure prior to this prescription 1) Prescribing measures (e.g., dosage and days' supply) are calculated and reported for patients not on LTOT using linked prescriptions; 2) Linked prescriptions are flagged as prescriptions for postsurgical pain EXCLUDE these flagged prescriptions from the remainder of the linkage algorithm (i.e., nonsurgical acute pain, chronic pain, and SCD pain) Link nonexcluded opioid prescriptions to visits, using the following criteria: 1) Prescription filled within 7 days of a visit, including the visit day; 2) No visit with any other pain indication in the study between the linked visit and the date on which the prescription is filled 1) For each patient, identify the first claim of a specific nonsurgical acute pain ICD-10-CM code, i.e. "index diagnosis"; 2) Apply the criteria for defining nonsurgical acute pain, i.e. pain lasts less than 90 days after the index diagnosis AND no same diagnosis in 6 months prior to the index diagnosis EXCLUDE these flagged prescriptions from the remainder of the linkage algorithm (i.e., chronic pain and SCD pain)         45.0 (28.0, 75.0) a Reported outcome data (prescribing rate, MME, and days' supply) reflect prescriptions supplied for a specific procedure or visit, meaning that the prescribing rate is anchored to visits/procedures. b Long-term opioid therapy (LTOT) is defined as 1) having at least three opioid prescriptions in the three months prior to the index visit; 2) more than 60 total days of opioid supply; and 3) the gap between the end of one prescription and the next prescription was fewer than 10 days. c Data from the OptumLabs Data Warehouse, 2017. d Data from the Marketscan Multi-State Medicaid Database, Q4 2016-Q3 2017. e Due to the small-cell suppression policy of OLDW, this is the largest allowable percentage we can report.

eTable 3. Opioid Prescribing Rates and Amounts a for Postsurgical Pain Management Among Patients on Long-term Opioid Therapy b in the United States, by Indication and Insurance Type, 2017
Abbreviations: LTOT = long-term opioid therapy; Rx = prescription; MME = morphine milligram equivalents (available at https://www.cdc.gov/drugoverdose/resources/data.html). (Q1, Q3) represent the interquartile range.  c Ages 0-18 years are not included here because of too little data to report.

eTable 4. Opioid Prescribing Rates and Amounts a for Nonsurgical Acute Pain Among Privately Insured Patients Not on Long-term Opioid Therapy b in the United States, by Indication and Age
d Patients not on long-term opioid therapy (LTOT) were those whose prescriptions did not meet the LTOT criteria.
e Long-term opioid therapy (LTOT) is defined as 1) having at least three opioid prescriptions in the three months prior to the index visit; 2) more than 60 total days of opioid supply; and 3) the gap between the end of one prescription and the next prescription was fewer than 10 days.