Data

To address some well-known limitations of existing data, we took a nuanced approach to calculating a more accurate true prevalence of unhealthy opioid use – as we did in our previous analysis of Virginia’s 2021 opioid data.

Fatal Opioid Overdoses

Estimates of the number of overdose deaths in 2023 rely on data from the Virginia Department of Health Forensic Epidemiology Fatal Overdose Tables – by locality and year.7 Data were collected from the Virginia Department of Health (VDH) website, and data on “All Opioid” related deaths were included in the analysis.8 VDH’s Office of the Chief Medical Examiner investigates and certifies the cause of death on all drug overdose death events that occur in Virginia. The definition of an opioid overdose death is designed to match as closely as possible the categorization from the National Vital Statistics System data on Drug Overdose Death Counts, and includes drug categories of heroin, prescription opioids, synthetic drugs like fentanyl, and other unspecified narcotics9:

Drug overdose deaths are identified using underlying cause-of-death codes from the Tenth Revision of ICD (ICD–10): X40–X44 (unintentional), X60–X64 (suicide), X85 (homicide), and Y10–Y14 (undetermined). Drug overdose deaths involving selected drug categories are identified by specific multiple cause-of-death codes. Drug categories presented include: heroin (T40.1); natural opioid analgesics, including morphine and codeine, and semisynthetic opioids, including drugs such as oxycodone, hydrocodone, hydromorphone, and oxymorphone (T40.2); methadone, a synthetic opioid (T40.3); synthetic opioid analgesics other than methadone, including drugs such as fentanyl and tramadol (T40.4); cocaine (T40.5); and psychostimulants with abuse potential, which includes methamphetamine (T43.6).

Opioid overdose deaths are identified by the presence of any of the following MCOD codes: opium (T40.0); heroin (T40.1); natural opioid analgesics (T40.2); methadone (T40.3); synthetic opioid analgesics other than methadone (T40.4); or other and unspecified narcotics (T40.6).

To adjust for missing data on death certificates and in the state mortality data, all county death counts are inflated by two percent. Virginia’s adjustment is very minor compared to other states, see Ruhm (2017) Geographic Variation in Opioid and Heroin Involved Drug Poisoning Mortality Rates for an explanation and detailed data on adjustment rates by state.10

We estimated the age-distribution of deaths in each county and independent city based on the single-year age distribution of opioid overdose deaths for the entire Commonwealth for the ages most commonly accounting for opioid deaths (ages 21 to 62) and then from the age distribution for the entire United States, rescaled to match the average Virginia rate, for all other ages. Supplemental population data were required for this approach and were taken from the US Census National and State Population Estimates for 2023.11 (The age distribution of opioid deaths in each county is required to estimate the total lost future earnings for each fatality.)

Nonfatal Unhealthy Opioid Use (UOU)

The counts of opioid use disorders were previously estimated using data from the National Survey on Drug Use and Health (NSDUH) alone, but our approaches were updated in our 2021 analysis, and this updated approach was repeated for 2023.12

For this year’s data update, we incorporate findings from Keyes et al. that NSDUH data may undercount the true population prevalence of UOU.13 As a result, we use an approach described in Section 2.4 of that paper that estimates UOU counts based on a multiplier of opioid mortality counts, by locality. This multiplier takes into account changes in the prevalence of more deadly, synthetic opioid use over time, and we use data collected from the 12 months of 2023 in Virginia from CDC Provisional Drug Overdose Deaths to estimate the share of opioid deaths in Virginia that occurred due to the use of synthetic opioids such as fentanyl.14 For the few counties and localities that had “0” overdose deaths (and by this approach would have estimated “0” individuals with UOU), we instead use the prior method of estimating county-level UOU rates based on the statewide estimate from NSDUH and the county’s share of ED Opioid Overdose visits from VDH data, described below.15

Our current approach substantially increased the estimates of UOU from our initial calculations for 2020. For data from NSDUH, estimates of UOU are higher in 2021 and 2023 because 2021 was the first year Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM-5) criteria were used to estimate rates and also because, “beginning with the 2021 NSDUH, questions on prescription drug use disorder were asked of all past year users of prescription drugs, regardless of whether they misused prescription drugs,” increasing the overall estimated rates of UOU.16

Further, our departure in using the Keyes et al. work to estimate county-level UOU rates further increases estimated rates as this approach:

• attempts to correct for undercounting of UOU in NSDUH data using multiplier approaches
• allows for greater county-level specificity of UOU rates, based on actual county-level mortality data (whereas prior data apportioned state-level UOU count estimates based on state population)

Opioid-Related Health Conditions

Injecting drugs such as opioids can place people who use them at risk for becoming infected with blood-borne diseases. If a person becomes pregnant while using, their child’s health may be impacted, as well. As with other opioid-related health issues, harm reduction and evidence-based care can improve outcomes.

Neonatal Abstinence Syndrome (NAS)

NAS is a suite of birth and development complications due to in-utero opioid exposure. Babies born with NAS were identified using locality-level data published by VDH.

HIV, Hepatitis C, Hepatitis B, and Tuberculosis

Primarily, people who inject drugs (including opioids) are at risk for contracting HIV, hepatitis C (HCV), hepatitis B (HBV), and tuberculosis. National estimates suggest 9.4% of new HIV cases17 and 60% of new HCV cases are linked to injection drug use.18

To calculate increased risk of infections related to opioid use, we used estimates of injection drug use rates among people with unhealthy opioid use by locality.19 Having estimated the population of opioid injection drug users by locality, we applied the probabilities and costs gathered by Jiang et al. to model health care utilization related to HIV, HCV, HBV, and tuberculosis among people who inject drugs at the locality level. It is important to note this is an update from pre-2021 analyses that used heroin use as a proxy for injection drug use and previously used estimates of heroin drug use from NSDUH. We believe this new approach better estimates the true risks of contracting HIV, HCV, HBV, and tuberculosis that people who use opioids face in a changing opioid epidemic, with the trends of:

• the declining prevalence of heroin use alone
• alongside an increasing prevalence of other synthetic opioids)

Notably, this change increases the expected rates of these conditions beyond the increase already expected due to higher UOU rate estimates in the 2021 and 2023 findings.

Locality-level data on new HIV20, HCV21, HBV22, and tuberculosis23 diagnoses were downloaded from the VDH website. We used observed new cases to adjust the distribution of the total costs among localities without changing the Virginia total. This helps reflect that areas with higher prevalence of either disease will bear disproportionately high costs due to greater likelihood of contagion, locally.

Lost Labor

Lost productivity represents productivity lost due to fatal opioid overdoses, non-fatal opioid use disorder, and opioid-related incarceration, combined with collective average lifetime earnings of the people who had these experiences in 2023. (This prediction was based on Virginia residents of similar ages and adjusted to each Virginia locality using data on the median county income for an individual with a high-school diploma relative to the state and national average of this statistic.)

Losses Due to Fatal Opioid Overdose

To estimate the productivity losses from each opioid overdose fatality, we apply methods used in Altarum's Value of Health modeling6 using data on:

• earnings
• tax rates
• non-opioid related mortality rates (taken from the CDC WONDER mortality data)

Estimated effective tax rates are taken from outputs of the National Bureau of Economic Research’s TAXSIM model for the state of Virginia and are applied evenly to each county and individual independent city.24

Lost future earnings by age are estimated for each county by:

• applying individual annual earnings by age from the American Community Survey (2017) 1-year microdata25
• smoothing them using a 5-year moving average
• adjusting them to each Virginia county using data on the median county income for an individual with a high-school diploma relative to the state and national average of this statistic for 2023. The median county income was again based on American Community Survey data26

To estimate the 2023 economic impact of lost future productivity, we apply the following assumptions to the earnings data:

• a 1% annual real growth rate in future earnings for all individuals
• a 3% annual discount rate to all earnings in future years

Losses Due to Non-Fatal Opioid Use Disorder

To estimate the productivity losses due to non-fatal UOU, we applied data from prior research on the impact of substance use disorders on labor force participation and wages and the county level earnings data from above.27

Losses Due to Opioid-Related Incarceration

The final component of opioid productivity impacts is estimated for the population of state residents incarcerated due to drug crimes related to opioids. While small, this component estimates the impact of lost workforce participation for 2023 from those incarcerated. We compute these with the following approach:

• Estimates of the state incarcerated population are taken from: the Bureau of Justice Statistics’ “Prisoners in 2021” report and the number of those imprisoned for drug crimes is estimated by applying the national percentage of inmates incarcerated for drug-related crimes to the total state incarcerated population.28
• The percentage of these drug crimes related to opioids was estimated using the ratio of estimated opioid drug arrests to total drug arrests in the Commonwealth, taken from the 2023 data from the Virginia Department of Criminal Justice Services (DCJS).29 Because the DCJS data only report on drug arrests by the “primary drug type”, we estimate the ratio of opioid associated arrests to all arrests incorporating all opioid arrests in the numerator, plus 50% of simulant and barbiturate arrests, due to the increasing prevalence of opioids in these other drug supplies.30

Total earnings lost are estimated using the median earnings data from above.

Our estimates of lost productivity due to incarceration do not include the costs of criminal justice or government contributions to prison costs but are solely the lost potential earnings of individuals who have been incarcerated.

Health Care Costs

Health care costs directly related to opioid use included responses to overdoses, hospital stays, and emergency department (ED) visits. The majority of health-care costs, though, were from care for conditions indirectly related to opioid use, such as neonatal abstinence symptoms in babies, hepatitis B, hepatitis C, HIV, and tuberculosis. We detail each category below.

Costs for Health Care Directly Related to Opioid Use

The total direct health care cost in each locality was estimated as the sum of costs from:

• emergency department visits for:
  • opioid overdose
  • other opioid-related health concerns
• inpatient stays:
  • opioid overdose
  • other opioid-related health concerns

For these four categories of events, Altarum:

• estimated hospital, ambulance, and naloxone costs
• quantified and monetized health-care utilization by locality

To do this, Altarum used the best-available imputation methods to address missing data based on federal-, state-, and local-level evidence. To estimate the cost of each of the health care use cases, a brief literature scan was conducted through May 2020 to identify the most recent estimates of opioid health care costs, such as costs of an overdose hospitalization. Altarum’s Health Spending Economic Indicators were used to adjust costs and payments for inflation over time where necessary.31

Hospitalizations

For each Virginia locality, Altarum estimated the number of opioid-attributable inpatient hospital stays for opioid-related acute health complications. Costs and payments associated with these stays were estimated by locality, as well. Commonwealth data provided by the Virginia Department of Health on overdose hospitalizations were first incorporated to generate estimates and complemented by federal data to identify non-overdose opioid stays.

Estimates of the number of overdose hospitalizations in 2023 are based on Healthcare Cost and Utilization Project hospitalizations and opioid-related discharges for Virginia.32 Because 2023 data were not yet available, we:

• adjusted 2022 data by the fraction by which opioid ED visits increased from 2022 to 2023
• apportioned this state-level estimate of overdose hospitalizations and discharges among localities in proportion to their relative number of opioid overdose ED visits
• computed costs using national estimates of opioid-poisoning inpatient stay costs provided by Inocencio et al. (2013)33

Emergency Department Visits

Overdose-related ED visit numbers were counted from the VDH’s public Emergency Department Visits for Unintentional Drug Overdose Among Virginia Residents statistics, which provided annual counts for ED visits for opioid overdoses by the patient’s locality of residence.34 For overdose ED costs:

• The average cost per visit was applied from estimates by Inocencio et al. (2013), which Altarum adjusted for national inflation in hospital outpatient care.33
• Payments were estimated by applying a payment-to-cost ratio, already calculated for overdose inpatient stays, to the estimated ED cost.

We also included non-overdose ED visits related to other opioid health concerns; specifically, we included “treat and release” (T&R) visits, or ED visits that did not result in hospitalization. We estimated non-overdose T&R visits for each locality by:

• multiplying each locality’s estimated opioid inpatient stays by the national ratio of opioid-related ED T&R visits to inpatient stays reported in the Fast Stats tables that estimated sub-state inpatient stays (using 2022 data in the absence of 2023 updates)35
• approximating the number of overdose visits that were T&R
• assuming all overdose inpatient stays began as an ED visit and subtracting them from the surveilled ED visits for overdose

We estimated their cost by:

• applying costs from Inocencio et al. (2013) and adjusting for inflation to get total costs per locality33
• multiplying these ED T&R costs by the payment-to-cost ratio already calculated for overdose inpatient stays

Ambulance and Naloxone Costs

Altarum calculated ambulance costs for ED visits for:

• overdose
• T&R visits for non-overdose
• non-overdose inpatient stays

(Overdose inpatient stays were assumed to be captured by the overdose ED visit data, whereas non-overdose cases could only be ED T&R or inpatient, but not both.) We multiplied our estimate of non-overdose inpatient stays by 71%, the share of opioid-related inpatient stays that originated as ED visits in the Mid-Atlantic region.36 The overdose ED, non-overdose ED T&R, and non-overdose inpatient stays are assumed to provide mutually exclusive, comprehensively exhaustive groups of ED cases, any of which may have involved an ambulance call.

Ambulances were assumed to be used in 75% of all ED cases, and ambulance costs were taken from Inocencio et al. (2013) national estimates, adjusted for inflation in general hospital-care prices from 2011 to the year of study.33 One naloxone dose was assumed used in all ambulance cases for overdose, multiplied by an estimated average $70 per dose.

Costs for Health Care Indirectly Related to Opioid Use

Altarum estimated the costs of the following health conditions attributable to opioid use: neonatal abstinence syndrome, HIV, hepatitis C and B viruses, and tuberculosis (diseases modeled by Jiang et al. (2017).38 Our estimates are based on the following data:

• the prevalence of opioid use disorder
• injection drug use
• VDH surveillance data of new cases of health conditions with opioid use as a known risk factor

Neonatal Abstinence Syndrome

Costs associated with birth were applied from Corr & Hollenbeak (2017)38, while non-birth costs associated with the first 8 years of life were taken by applying Liu et al.’s (2019) NAS-attributable spending multipliers to national health spending per capita for children ages 0-8.39 Birth and postpartum costs were separated out using estimates from Bui et al. (2017).40 Furthermore, Altarum used age-specific spending data for 2016 from the Institute for Health Metrics and Evaluation Disease Expenditure project database (received by request from Joseph Dieleman, Ph.D.)41, and Child Trends population estimates by age for 2016 gave a denominator for per capita costs.42

HIV, HCV, HBV, and Tuberculosis

HIV, HCV, HBV, and tuberculosis all have lasting effects that require high medical costs to treat symptoms upon their onset. These contagious diseases spread via injection drug use, such as with heroin. We used the model provided by Jiang et al. (2017)37, and incorporated estimates of injection drug use19 rates among people with unhealthy opioid use, to estimate the state’s total future healthcare costs owing to new injection drug use infections during 2023. Having estimated the population of opioid injection drug users by locality, we applied the probabilities and costs gathered by Jiang et al. to model HIV, HCV, HBV, and tuberculosis costs in that injection drug-user group at the locality level. It is important to note this is an update from pre-2021 analyses that used heroin use as a proxy for injection drug use and previously used estimates of heroin drug use from NSDUH. As a result of the changing opioid epidemic (and the declining prevalence of heroin use alone, alongside an increasing prevalence of other synthetic opioids), we believe this new approach better estimates true risks to HIV, HCV, HBV, and tuberculosis. This change increases the expected rates of these conditions beyond the increase already expected due to higher UOU rate estimates in the 2021 and 2023 findings.

For HIV, HCV, HBV, and tuberculosis we used observed new cases to adjust the distribution of the total costs among localities without changing the Virginia total. This helps reflect that areas with higher prevalence of either disease will bear disproportionately high costs due to greater likelihood of contagion, locally.

Locality-level data on HIV, HCV, HBV, and tuberculosis new diagnoses were downloaded from the VDH website.43 Using the costs in Jiang et al.’s model37, these new diagnoses were monetized; national estimates suggest 9.4 percent of new HIV cases17 and 60% of new HCV cases are linked to injection drug use18, so these shares of the disease costs were applied to each locality to estimate the opioid-attributable cost from documented new cases. The remainder of the Virginia cost of illness from HIV, HCV, HBV, and tuberculosis, as modeled based on Jiang et al., were then distributed among localities according to the size of localities’ estimated injection drug user populations. Where applicable, costs of treatment were adjusted by Altarum’s all-item measure of health care price inflation to reflect 2023 dollars.

Child Welfare Services

Ideally, parents or other caregivers pursuing recovery from opioid use would not be separated from their children – there are many benefits to keeping substance use-affected families together, when safe and appropriate to do so.44 However, opioid use can impact children’s ability to live safely in their homes – whether a parent or caregiver is actively using (and potentially exposing children to opioids), has been incarcerated for opioid-related crimes, or has died from a fatal overdose. In some cases, children accidentally ingest opioids or intentionally begin taking them, either of which can be dangerous to their health.

Virginia provides a range of child welfare services for families impacted by opioid use. They may evaluate safety risks to children in the home, work to prevent abuse or neglect, and if needed, place a child in foster care, kinship care45, or a different environment.

To estimate costs to Virginia’s Department of Family Services, we use data from a 2019 paper in the American Journal of Managed Care, by Crowley et al.46 We apply their estimates for the 2016 costs nationally to the child welfare system (including Child Protective Services, in-home services, and foster care services) to our own estimates of the national case count of UOU in that year using the Keyes et al. approach.13 This is used to compute an estimated cost per UOU case to the child welfare system (inflated to 2023 dollars), which we apply to the 2023 counts of UOU in Virginia to estimate a statewide cost to the Department of Family Services.

To estimate the cost impact per county in 2023, we:

• use data on the number of child referrals to Child Welfare Services by locality from 2023 data provided by the Virginia Department of Social Services49
• apportion the total state cost to each locality, based on each locality’s share of the overall number of referrals across the state

Note: For storytelling purposes, we combined Child Services with K–12 Education in our primary data visualizations, as both relate to impacts on children. However, our raw data set includes separate numbers for each category.

K–12 Education

Children impacted by opioid use may need trauma-informed counseling, academic support, and other services at school.48

We estimated additional direct expenditures on Virginia’s K-12 education system due to UOU using methods from previous work by the National Center on Addiction and Substance Abuse that estimated the national cost to the K-12 education system due to substance use disorders.49 We estimated Virginia’s annual opioid-related costs to the public K–12 education system by:

• dividing the national total value by data on the total number of substance use disorder conditions in 2016 from NSDUH data, and then inflated the number to 2023 dollars
• applying to the locality-level data on the number of individuals with UOU in 2023 for Virginia

Criminal Justice

Criminal justice-related costs represent spending related to opioid-related arrests and incarcerations, as well as the need for the courts and police protection that accompany them.

Annual costs for opioid-related arrests by locality are based on:

• county-level drug arrest rates data provided by the Virginia State Police50
• the methods described in the Productivity Losses Due to Incarceration section above that determined the number of opioid-related arrests in 2023
• estimating cost per arrest accruing from combined police protection and court system costs, based on a per-drug-related arrest cost value from a 2013 Justice Policy Institute analysis, adjusted for inflation to 2023 dollars51

Costs for opioid-related incarceration by locality are based on:

• a 2012 Vera Institute of Justice report on the average cost per inmate in Virginia52
• applied to the estimated number of opioid-related incarcerations in 2023

The following sections explain how we divided economic costs within our data set by payer. For details on missing human, social, and economic costs, see our data limitations.

Note: We refer to “taxpayers” and “governments” interchangeably, recognizing that due to the opioid crisis:

• Taxpayer-funded programs face increased costs related to health care, child welfare services, K–12 education, and criminal justice.
• Governments collect less tax revenue than they would with a working-age population with a longer healthy lifespan.

Families and Businesses

Our cost estimates to families and businesses include both lost labor and health care – although we have only broken out lost labor costs in the raw data set. We attributed the costs as follows:

• We apportioned lost labor costs using the tax rate data described in the earlier section on lost labor costs and Altarum Value of Health tool methods.6
• We apportioned health care costs based on VDH data we requested on the breakdown of overdose charges by payer and the proportion of coverage type paid for by each major insurer.
  • Private insurance costs were attributed to families and businesses.

Local and State Taxpayers

Our cost estimates to state and local taxpayers include all sectors in our data model.

• We apportioned lost labor costs using the tax rate data described in the earlier section on lost labor costs and Altarum Value of Health tool methods.6
• We apportioned health care costs based on VDH data we requested on the breakdown of overdose charges by payer and the proportion of coverage type paid for by each major insurer.
  • Medicaid costs attributions were split between the state and federal government based on the fiscal year Federal Medical Assistance Percentage (FMAP) for 2023.53
• We attributed all costs related to child welfare services, K-12 education, and criminal justice to the state government (not the federal government, due to limited available data). We localized costs incurred by each city and county as described in the methods for each sector.

Federal Taxpayers

Our cost estimates to federal taxpayers include only lost labor and health care. (Due to limited data, we were not able to include costs related to children or criminal justice.) We attributed the costs as follows:

• We apportioned lost labor costs using the tax rate data described in the earlier section on lost labor costs and Altarum Value of Health tool methods.6
• We apportioned health care costs based on VDH data we requested on the breakdown of overdose charges by payer and the proportion of coverage type paid for by each major insurer.
  • Medicaid costs were attributed between the state and federal government based on the fiscal year FMAP for 2023.53
  • Medicare costs were attributed to the federal government.