The measure specification below is for dizziness and stroke in the Emergency Department (ED). Dizziness is the stroke symptom most likely to be misdiagnosed, generally as a benign inner ear condition. This measure seeks to identify patients with dizziness or vertigo symptoms whose strokes are missed at an initial (index) ED visit and discovered at a later (outcome) hospitalization. The “Avoid H.A.R.M. Dizzy-Stroke-ED” measure specification is provided below.

Contact us at dxcenter@jhu.edu.

Measure Specifications

sp.13) State the numerator. Brief, narrative description of the measure focus or what is being measured about the target population, i.e., cases from the target population with the target process, condition, event, or outcome).

DO NOT include the rationale for the measure.

The number of Emergency Department treat and release index visits during the performance period that are followed within 30 days by an inpatient hospital admission to any hospital that ends in a primary diagnosis of stroke.

sp.14) Provide details needed to calculate the numerator. All information required to identify and calculate the cases from the target population with the target process, condition, event, or outcome such as definitions, time period for data collection, specific data collection items/responses, code/value sets.

Note: lists of individual codes with descriptors that exceed 1 page should be provided in an Excel or csv file in required format at sp.11.

For each patient’s index ED visit identified in the denominator, identify if the patient had an inpatient hospital admission to any hospital within 30 days of their ED discharge date that resulted in a primary diagnosis of stroke. The ICD-10 codes to be used to identify patients with a primary diagnosis of stroke can be found in the submitted Excel file.

sp.15) State the denominator. Brief, narrative description of the target population being measured.

Patients discharged from the ED with a primary diagnosis code of “benign dizziness”. A patient’s first such discharge during the performance period will be considered the “index visit”. Any subsequent ED discharge with a diagnosis of ‘benign dizziness’ that falls outside a 360-day follow-up window from the previous qualifying “index visit” will be considered another distinct ‘index visit’.

sp.16) Provide details needed to calculate the denominator. All information required to identify and calculate the target population/denominator such as definitions, time period for data collection, specific data collection items/responses, code/value sets.

Note: lists of individual codes with descriptors that exceed 1 page should be provided in an Excel or csv file in required format at sp.11.

Using a 36- month performance period, identify those ED patients who were discharged from the ED with a primary diagnosis of “benign dizziness”. This includes patients with either with (1) a specific benign dizziness diagnosis (e.g., benign paroxysmal positional vertigo) or (2) a non-specific, symptom-only dizziness diagnosis (i.e., dizziness or vertigo, not otherwise specified). The ICD-10 codes to be used to identify patients with a primary diagnosis of “benign dizziness” can be found in the submitted Excel file.

A patient’s first ED discharge during the performance period meeting the above criteria should be included in the denominator. This patient discharge is considered the patient’s first “index visit”. A patient’s second “index visit” is the first subsequent ED discharge meeting the above criteria that is more than 360 days after the first index visit’s ED discharge date and this “index visit” should also be included in the denominator. A patient’s third “index visit” is the first subsequent ED discharge meeting the above criteria that is more than 360 days after the second index visit’s ED discharge date and this “index visit” should be included in the denominator. A patient’s fourth “index visit” is the first subsequent ED discharge meeting the above criteria that is more than 360 days after the third index visit’s ED discharge date and this “index visit” should be included in the denominator.

The denominator value is the count of the number of ED “index visits” with a primary discharge diagnosis of “benign dizziness” during the performance period. The maximum number of “index visits” for a single patient in a 36-month performance period is 4.

sp.17) Describe the denominator exclusions. Brief narrative description of exclusions from the target population.

The measure has no exclusions. All patients discharged from the ED with "benign dizziness" as their primary diagnosis code are included in the measure denominator.

sp.24) Diagram or describe the calculation of the measure score as an ordered sequence of steps. Identify the target population; exclusions; cases meeting the target process, condition, event, or outcome; time period of data, aggregating data; risk adjustment; etc.

Steps to calculate an ED’s risk of misdiagnosed-related harm from missed stroke.

1. Step 1 – Identify all patients discharged from the ED with a primary diagnosis of “benign dizziness” during the 36-month performance period.
2. Step 2 – A patient’s first ED discharge during the 36-month performance period with a primary diagnosis of “benign dizziness” should be included in the denominator. This patient discharge is considered the patient’s first “index visit”. A patient’s (potential) second “index visit” is the first subsequent ED discharge with a diagnosis of “benign dizziness” that is more than 360 days after the first index visit’s ED discharge date. A patient’s (potential) third “index visit” is the first subsequent ED discharge with a diagnosis of “benign dizziness” that is more than 360 days after the second index visit’s ED discharge date. A patient’s (potential) fourth “index visit” is the first subsequent ED discharge with a diagnosis of “benign dizziness” that is more than 360 days after the third index visit’s ED discharge date. Index visits that do not have patients enrolled for at least 360 days after the index visit were excluded.
3. Step 3 – Count the number of ED “index visits”—this is the denominator value. The maximum number of “index visits” for a single patient in a 36-month performance period is 4.
   “Observed” Rate Calculation
4. Step 4 – For each “index visit” in Step 3, identify if the patient had an inpatient admission to any hospital within 30 days of their ED discharge that resulted in a primary diagnosis of stroke. Count the number of “index visits” that meet this criterion—this is the short-term 30-day numerator value.
5. Step 5 – Measure the observed rate. Crude short-term 30-day rate per 10,000 visits = (Step 4: [number of short-term stroke hospitalizations within 30d + alpha] / Step 3: [number of eligible ED benign dizziness discharges in the performance period + 1) x 10,000. The constants “alpha” = 1/1,000 (for the numerator) and “1” (for the denominator) are added to avoid issues with possible zero counts [see footnote “*” below for clarification].
   “Expected” Rate Calculation
6. Step 6 – For each “index visit” in Step 3, identify if the patient had an inpatient admission to any hospital with a primary diagnosis of stroke in the time window 91 days through 360 days following their ED index visit discharge. Count the number of “index visits” that meet this criterion.
7. Step 7 – Divide the number of 91 days through 360 days strokes identified in Step 6 by 9 to calculate the average delayed rate per 30 days to obtain a ‘monthly’ value. This is the long-term 30-day numerator value.
8. Step 8 – Measure the expected rate. Crude long-term 30-day rate per 10,000 visits = (Step 7: [average number of long-term stroke hospitalizations per 30d + alpha] / Step 4: [number of eligible ED benign dizziness discharges in the performance period who did not experience a stroke in the prior 90 days + 1 - (3 x alpha)]) x 10,000. The denominator should exclude those patients who experienced a stroke prior to 90 days as we are only counting the first stroke in the 91-360 days post index visit. The constants “alpha” = 1/1,000 (for the numerator) and “1 - (3 x alpha)” (for the denominator) are added to avoid issues with possible zero counts [see footnote “*” below for clarification].
   “Attributable” Rate (Measure) Calculation
9. Step 9 – Attributable 30d rate per 10,000 visits = Step 5 (crude short-term 30d rate) – Step 8 (crude long-term 30d rate)

* The constants “alpha” = 1/1,000 (for the numerator) and “1” (for the denominator) are added to avoid issues with possible zero counts. This is equivalent to a posterior estimation using Beta (alpha, 1-alpha) as prior for each 30-day rate. It is similar to the "add 0.5" approach in the Fisher´s exact test with low counts, except that here, the 30-day stroke return rate of alpha = 1/1,000 is used as prior as opposed to 1/2 as in the Fisher´s exact test. This prior translates to adding 1 observation with a 30-day stroke return rate of alpha when calculating the observed 30-day rate and the expected 30-day rate. The estimation is asymptotically unbiased and consistent. The effect of this statistical adjustment is negligible but penalizes the measure towards no harm. The statistical adjustment factor (alpha) of 1/1,000 was chosen to be similar to the long-term, baseline stroke risk after ED discharge and is reasonable based on our current data and that from prior research studies. Removing “3 x alpha” from the denominator in calculating the expected 30d rate is due to having to remove patients that already experienced a stroke hospitalization prior to 90d.

Download H.A.R.M ICD Codes Excel File

Risk Model Details

If an outcome or resource use measure is not risk-adjusted or stratified, provide rationale and analyses to demonstrate that controlling for differences in patient characteristics (i.e., case mix) is not needed to achieve fair comparisons across measured entities.

Our measure uses a statistical risk difference approach (observed [short-term stroke risk] minus expected [long-term/baseline stroke risk]) using the same patient cohort. As a result, controlling for differences in patients characteristics (case mix) is not needed to achieve fair comparisons across entities.

Risk Difference Approach: The risk-difference measure is a difference between two rates (observed minus expected), reflecting the observed stroke events in the first 30 days after an ED discharge (i.e., are likely to represent more than a chance association between the ED discharge and inpatient admission). above the expected epidemiologic base rate This approach accounts for inter-institutional differences in the underlying stroke risk of their specific patient populations including any social determinants of health in the affected population. It represents a conservative estimate of the rate of misdiagnosis-related harms from missed stroke because it assumes that long-term strokes (e.g., 91-360 days post discharge) are not likely to be preventable harms linked back to the original misdiagnosis.

Risk Difference Parameters: The short-term observed rate is measured as the number of stroke hospitalizations per 10,000 discharges in the first 30 days and is called the short-term 30-day rate of stroke hospitalization. The short-term expected rate is estimated in the exact same patients by taking the average 30-day rate of stroke admission during a long-term outcome assessment window. The long-term window (91 days to 360 days post discharge) is chosen to reflect the epidemiologic base rate of stroke (i.e., after the short-term risk of a misdiagnosis leading to preventable major stroke has definitively passed). The stroke rate per 30-day period during this long-term 270-day window is obtained by dividing the numerator by nine and is called the long-term 30-day rate.

Risk Difference Rationale: Patients that have stroke hospitalizations within 30 days of an ED “benign dizziness” discharge represent patients that are misdiagnosed at the ED index visit, but also include some patients that are not misdiagnosed (i.e., do, in fact, have benign dizziness) who go on have a coincidental stroke event due to baseline (biological/sociocultural) stroke risk. This baseline stroke risk is reflected by the long-term population-specific stroke rate which is not related to the institutional rate of misdiagnosis or short-term harms (i.e., 30-day stroke admissions). This relationship is most evident when viewed as a longitudinal incidence rate curve for stroke hospitalization (Fig. 1). This curve matches the natural history/biological profile of major stroke following minor stroke and TIA (Fig. 2A/B).

Figure 1. Weekly incidence rate curve of stroke hospitalization post ED discharge as “benign dizziness.” KPMAS data from the performance period from 2010-2014 at all outpatient sites (ED, ambulatory care). Data reflect 56,746 treat-and-release visits for “benign dizziness.” Shown in black are stroke hospitalizations, and shown in red are heart attack hospitalizations (gray shading represents 95% confidence intervals for each). Early returns for stroke hospitalization above the epidemiologic base rate in the first few weeks after discharge reflect potentially preventable harms from stroke missed at the index visit. The comparison outcome of heart attack demonstrates the association is specific.

 This risk difference approach uses an institution-specific delayed (91d–360d) stroke hospitalization rate to approximate the baseline short-term stroke risk for the population in question. This long-term window is chosen because, biologically-speaking, the short-term risk of major stroke after minor stroke or TIA levels off by approximately 30 days after the initial cerebrovascular event (Figure 2B). By using the risk difference, the measure quantifies only the “excess” short-term stroke rate (attributable risk) due to misdiagnosis above the base rate for the population in question. Thus, the risk difference accounts for all relevant demographic differences across populations including biological and social and determinants of health that may lead to population-level variation in baseline stroke risk.

Rationale for No Demographic Risk Adjustments: Other racial or demographic disparities in institution-specific risk of misdiagnosis that are linked to the institution-specific patient populations should be measured appropriately rather than “adjusted” away (e.g., racial bias, racial minorities are at higher risk of being misdiagnosed [Newman-Toker et al., Diagnosis, 2014; PMID: 28344918]).

Risk Difference Calculation: The risk difference calculation requires an observed and expected rate calculation. For each patient discharged from the ED with a “benign dizziness” diagnosis during the performance period, data on stroke hospitalizations must be available for a floating outcome assessment window of roughly 12 months (360 days). If stroke hospitalizations occur between post-ED day #1 and day #30 (i.e., mostly linked to misdiagnosis-related harms), they are counted in the numerator of the “short-term 30-day rate” (observed rate). If stroke hospitalizations occur between post-ED day #91 and day #360 (i.e., mostly linked to baseline biological or sociocultural stroke risk), they are counted in the numerator of the “long-term 30-day rate.” The long-term rate is normalized to a 30-day period equivalent rate over the 270-day outcome assessment window by dividing by nine (i.e., taking the average 30-day rate during those 270 days). A 270-day window is used for the average delayed 30-d rate calculation because of very low stroke base rates in this population (<0.1% [Newman-Toker, Ann Neurol, 2015; PMID: 26418192]); this increases the precision of the “expected” value.

• Crude short-term 30-day rate = {[number of stroke hospitalizations within 30d + alpha] / [number of eligible ED benign dizziness discharges in the performance period + 1]} x 10,000. This “short-term” rate includes the early peak rate (Fig. 1) of hospitalization after missed stroke and dominantly reflects misdiagnosis (but partly reflects the base rate). The measure is represented as the number of stroke hospitalizations per 10,000 benign dizziness discharges. The constants “alpha” = 1/1,000 and “1” are added to avoid issues with possible zero counts.
• Crude long-term 30-day rate = {([number of stroke hospitalizations from 91d-360d divided by 9] + alpha) / [number of eligible ED benign dizziness discharges in the performance period and no stroke diagnosis in the prior 90 days + 1 - (3 x alpha)]} x 10,000. This “long-term” rate approximates the epidemiologic “base” rate of stroke in the specific population in whom the short-term 30d rate is measured. The parameter is represented as the number of stroke hospitalizations per 10,000 benign dizziness discharges. The denominator should exclude those patients who experienced a stroke prior to 90 days since we are only counting the first stroke in the 360 days post index visit. The constants “alpha” = 1/1,000 and “1 - [3 x alpha]” are added to avoid issues with possible zero counts.
• Attributable short-term 30d rate = (crude short-term 30d rate) – (crude long-term 30d rate); the attributable short-term rate reflects the “excess” short-term (30d) rate of stroke above the base rate that is specific for the population in question. This is an estimate of the attributable risk of misdiagnosis-related harms from missed stroke. The parameter is represented as the number of stroke hospitalizations per 10,000 benign dizziness discharges.