Work Productivity and Activity Impairment (WPAI)

Work Productivity and Activity Impairment (WPAI)

captures how health affects paid work and regular activities over a 7-day recall period. The WPAI General Health (WPAI:GH) asks about employment, work time missed because of health, time missed for other reasons, time actually worked, impairment while working, and impairment in regular activities. Disease-specific WPAI versions retain the same scoring structure but attribute impairment to the named health problem. The derived domains are percentages from 0 to 100, where higher values mean more impairment.

Use the developer scoring rules, not a home-made sum

Health-related absenteeism is health-related hours missed divided by health-related missed hours plus hours actually worked. Presenteeism is the reported on-the-job impairment rating divided by 10. Overall work impairment combines absence and impairment during time worked: absenteeism + (1 - absenteeism) × presenteeism. Activity impairment is its separate 0-to-10 rating divided by 10 and applies whether or not someone is employed. Do not add the percentages: overall work impairment already accounts for both components.

Employment, denominator, and missing-item rules matter

Score each work domain from its own required inputs rather than discarding every work score because one item is absent. Absenteeism requires valid health-missed and worked hours with a positive denominator. Presenteeism requires a valid on-the-job impairment rating and known positive worked hours for a respondent eligible for work-domain assessment; a missing health-missed-hours response does not by itself erase that valid rating. Overall work impairment requires the component inputs, except that when all observed scheduled time was missed for health (health-missed hours > 0 and worked hours = 0), absenteeism and overall work impairment are 100% and presenteeism is not applicable. A respondent who is not employed has no work-domain score; never convert that to 0% impairment. Activity impairment remains independently scorable for employed and non-employed respondents. Preserve the instrument version, language, attribution (general health or condition-specific), recall period, completion date, and the reason a score is absent.

Measurement is not monetary valuation

WPAI reports perceived time and productivity impairment. Converting it to money requires additional choices: wage source, fringe benefits, employer replacement costs, labor-force participation, and whether to use a human-capital or friction-cost perspective. A 20% WPAI overall-work-impairment score is not automatically 20% of salary lost, and it should not be combined with claims-paid amounts without a stated perspective and double-counting assessment.

Pros, cons, and trade-offs

WPAI is brief, understandable, and can capture productivity impact that claims omit. Its cost is self-report and a short recall window; results may be affected by disease severity, occupation, benefit design, remote work, and cultural or language context. Relative to HRQoL measures that produce preference-weighted utilities, WPAI describes functional/productivity consequences rather than health utility. Relative to burden-of-disease-cost-of-illness, it supplies an input to indirect-cost estimation but does not by itself estimate societal cost.

When NOT to use — and when it is actively misleading

Do not use WPAI to infer productivity from claims, diagnosis codes, or employment status. Do not report a work-domain mean across all enrolled people while silently treating non-employed participants as zero impairment. Do not substitute a disease-specific WPAI score for a general-health score, or vice versa, without saying which attribution question was asked. Do not reproduce the proprietary questionnaire items or assume that a translated or modified instrument remains validated.

Data-source operational depth

• Survey / ePRO: The native source. Preserve raw permitted item responses, version/language, completion timestamp, and 7-day recall anchor; apply developer-authorized scoring only.
• EHR: EHRs may hold a completed questionnaire but generally cannot reconstruct valid WPAI domains from visit notes. Employment fields are not equivalent to the instrument's employment and time responses.
• Claims: Claims can measure direct medical spending but cannot observe absence, presenteeism, or activity impairment. Employer claims or disability feeds require separate validation and still do not produce WPAI without the questionnaire.
• Registry / linked data: Link survey/ePRO scores to clinical and claims data with explicit consent, date alignment, and linkage-availability reporting. The linked subset may differ in employment, severity, and response propensity.

Decision diagram

%%{init: {'themeVariables': {'fontSize': '16px'}, 'flowchart': {'nodeSpacing': 18, 'rankSpacing': 24}}}%%
flowchart TD
  A["Authorized version<br/>and 7-day recall"] --> B["Adjudicate employment<br/>Assess each domain separately"]
  B --> C["Absenteeism<br/>Health-missed / total work hours<br/>Positive denominator required"]
  C --> D["Presenteeism<br/>At-work rating / 10<br/>Positive worked hours required"]
  D --> E["Overall work impairment<br/>Combine available components<br/>All work missed for health: 100%"]
  E --> F["Activity impairment<br/>Activity rating / 10<br/>Employment not required"]
  F --> G["Report domain percentages<br/>and missingness separately"]

Index definitions

Worked example

Scenario

A 7-day disease-specific WPAI assessment is completed at baseline by four people in an RWE registry. The study will report health-related work impairment separately from activity impairment and will not monetize the score.

Dataset

Synthetic score inputs; numbers are permitted responses and not questionnaire text.

Steps

Result

The study reports work outcomes among score-eligible employed respondents and activity impairment among all valid respondents. It does not describe the non-employed group as having zero work impairment or turn the percentages into dollars without a separate valuation model.

Trade-offs

Code example

import math

def wpai_scores(employed, health_missed_hours, worked_hours,
                at_work_impairment_0_to_10, activity_impairment_0_to_10):
    def valid_nonnegative(value):
        return isinstance(value, (int, float)) and math.isfinite(value) and value >= 0

    def scale_0_10(value):
        return 100 * value / 10 if valid_nonnegative(value) and value <= 10 else None

    activity = scale_0_10(activity_impairment_0_to_10)
    eligible = employed is True
    valid_hours = valid_nonnegative(health_missed_hours) and valid_nonnegative(worked_hours)
    denom = health_missed_hours + worked_hours if valid_hours else None
    absenteeism = (100 * health_missed_hours / denom
                   if eligible and denom is not None and denom > 0 else None)
    presenteeism = (scale_0_10(at_work_impairment_0_to_10)
                    if eligible and valid_nonnegative(worked_hours) and worked_hours > 0 else None)
    if eligible and valid_hours and health_missed_hours > 0 and worked_hours == 0:
        overall = 100.0
    elif absenteeism is not None and presenteeism is not None:
        overall = absenteeism + (100 - absenteeism) * presenteeism / 100
    else:
        overall = None
    return {"absenteeism": absenteeism, "presenteeism": presenteeism,
            "overall_work_impairment": overall, "activity_impairment": activity}

synthetic_responses = [
    {"employed": True, "health_missed_hours": 4, "worked_hours": 28,
     "at_work_impairment_0_to_10": 5, "activity_impairment_0_to_10": 6},
    {"employed": True, "health_missed_hours": None, "worked_hours": 28,
     "at_work_impairment_0_to_10": 7, "activity_impairment_0_to_10": 8},
    {"employed": True, "health_missed_hours": 4, "worked_hours": 0,
     "at_work_impairment_0_to_10": None, "activity_impairment_0_to_10": 5},
    {"employed": False, "health_missed_hours": None, "worked_hours": None,
     "at_work_impairment_0_to_10": None, "activity_impairment_0_to_10": 4},
]
print([wpai_scores(**row) for row in synthetic_responses])

wpai_scores <- function(employed, health_missed_hours, worked_hours,
                        at_work_impairment_0_to_10, activity_impairment_0_to_10) {
  valid_nonnegative <- function(x) length(x) == 1L && is.finite(x) && x >= 0
  scale_0_10 <- function(x) if (!valid_nonnegative(x) || x > 10) NA_real_ else 100 * x / 10
  activity <- scale_0_10(activity_impairment_0_to_10)
  eligible <- isTRUE(employed)
  valid_hours <- valid_nonnegative(health_missed_hours) && valid_nonnegative(worked_hours)
  denom <- if (valid_hours) health_missed_hours + worked_hours else NA_real_
  absenteeism <- if (eligible && valid_hours && denom > 0) 100 * health_missed_hours / denom else NA_real_
  presenteeism <- if (eligible && valid_nonnegative(worked_hours) && worked_hours > 0) scale_0_10(at_work_impairment_0_to_10) else NA_real_
  overall <- if (eligible && valid_hours && health_missed_hours > 0 && worked_hours == 0) 100 else if (!is.na(absenteeism) && !is.na(presenteeism)) absenteeism + (100 - absenteeism) * presenteeism / 100 else NA_real_
  c(absenteeism=absenteeism, presenteeism=presenteeism,
    overall_work_impairment=overall,
    activity_impairment=activity)
}
synthetic_responses <- data.frame(
  employed = c(TRUE, TRUE, TRUE, FALSE), health_missed_hours = c(4, NA, 4, NA),
  worked_hours = c(28, 28, 0, NA), at_work_impairment_0_to_10 = c(5, 7, NA, NA),
  activity_impairment_0_to_10 = c(6, 8, 5, 4)
)
t(mapply(wpai_scores, synthetic_responses$employed, synthetic_responses$health_missed_hours,
         synthetic_responses$worked_hours, synthetic_responses$at_work_impairment_0_to_10,
         synthetic_responses$activity_impairment_0_to_10))

data wpai_scored;
  set wpai;
  if not missing(activity_impairment_0_to_10) and activity_impairment_0_to_10 >= 0 and activity_impairment_0_to_10 <= 10 then activity_impairment = 10*activity_impairment_0_to_10;
  else activity_impairment = .;
  absenteeism = .; presenteeism = .; overall_work_impairment = .;
  valid_hours = not missing(health_missed_hours) and not missing(worked_hours)
                and health_missed_hours >= 0 and worked_hours >= 0;
  if valid_hours then work_denom = health_missed_hours + worked_hours;
  else work_denom = .;
  if employed=1 and valid_hours and work_denom > 0 then absenteeism = 100*health_missed_hours/work_denom;
  if employed=1 and not missing(worked_hours) and worked_hours > 0
     and not missing(at_work_impairment_0_to_10) and at_work_impairment_0_to_10 >= 0
     and at_work_impairment_0_to_10 <= 10 then presenteeism = 10*at_work_impairment_0_to_10;
  if employed=1 and valid_hours and health_missed_hours > 0 and worked_hours=0 then overall_work_impairment = 100;
  else if not missing(absenteeism) and not missing(presenteeism) then do;
    overall_work_impairment = absenteeism + (100-absenteeism)*presenteeism/100;
  end;
run;

Citations

1. [1]Reilly MC, Zbrozek AS, Dukes EM. The validity and reproducibility of a work productivity and activity impairment instrument. Pharmacoeconomics. 1993;4(5):353-365.

1. [2]Walker-Bone K. The Work Productivity and Activity Impairment (WPAI) questionnaire. Occupational Medicine. 2026;76(2):92-93.

1. [3]HCV RWE PMOS Statistical Analysis Plan P15-405: Study P15-405 (OPALE), Version 1.0. 2018-02-15.