SAFTINet CERAsthma Protocol: Design & Methods

November 12, 2012

Aims & Hypotheses

• Aim: To model the association between a practice’s medical home characteristics and asthma outcomes in children and adults.

• H1: In both adults and children, greater practice-level “medical home” total and domain scores are associated with better asthma control.

• H2: Greater practice-level “coordinated asthma care” is associated with better asthma control.

Design• Prospective, longitudinal cohort study, utilizing survey

methodologies and secondary use of existing structured clinical, administrative, and claims data (henceforth “electronic health data”)

• Baseline period: January 1, 2011 through June 30, 2012– Determining baseline cohort membership, assignment to

practice, and baseline severity/risk for poor asthma control• Follow-up period: July 1, 2012 through December 31,

2013– Three consecutive 6-month blocks– Rolling cohort membership

Timeline

Concepts

• Asthma cohort membership• Exposure to medical home– Assignment to practice

• Surveillability– Eligibility for inclusion in analysis for a given block

Data Sources

• Practice surveys of medical home characteristics and general practice characteristics

• Structured electronic health data (including patient-reported outcomes) from the SAFTINet federated database – Outcomes:

• Asthma control test (*control category: good, poor, very poor)• Asthma exacerbations (at least one of: steroid bursts, cluster

of asthma encounters, inhaled beta-agonist administration, asthma-related hospitalization/ED visit)

Analytic Approach

• Linear mixed models– Patients (asthma outcomes) nested within practice

(medical home characteristics)• Hierarchical approach to variable entry into

model– Systematically rule out alternative explanations

(addressing bias)

Bias & Confounding• Selection bias

– Cohort membership/asthma diagnosis– Patient demographics– Asthma severity– Risk for uncontrolled asthma

• Misclassification bias– Incorrectly assigned to practice, exposure to MH– Incorrectly indicating absence/presence of exacerbation

• Information bias– Systematic differences in collection of outcomes data/documentation

(e.g., mode of data collection, increased access = more data)

DAG

Asthma Control (tn)-ACT (good, poor, very poor control)-Exacerbations 1+ of: -ED/hospital -Steroid burst -Az visit cluster -Inhaled beta-agonist

Medical Home (tn)-Personal clinician-Teams-Coordination/integration-Pt centeredness-QI-Organized/EBM-Access-Leadership-Registries/perf rep-Enhanced asthma care

Intermediate (tn)-Attention to asthma -More data on asthma-Provision of care -More access for in-house az tx-Use of spirometry-Patient adherence to tx-Provider guideline concordance-Continuity of care (1o care util)-Smoking status

Practice-level variables (to)-Practice population -Total unique pts, pt visits -Payer mix -Race/ethnicity overall mix -Total az patients

Practice-level variables (to)-Practice Size/Diversity -# providers (FTE) -# specialties in house-Practice location -MSA size (urban/rural)

Patient-level variables (to)-Patient demographics -Age -Race & Ethnicity -SES (family income, size) -Gender (maybe) -Urban/rural

Patient-level variables (to)-Risk for poor asthma control -Prior exacerbations -Prior utilization -Prior ACT scores-Asthma severity -Persistent/Intermittent (meds)

Patient-level variables (t1)-Risk for poor asthma control -comorbidities (allergy, rhinitis/sinusitis, GERD)

DAG

Asthma Control (tn)-ACT (good, poor, very poor control)-Exacerbations 1+ of: -ED/hospital -Steroid burst -Asthma visit cluster -Inhaled beta-agonist

Medical Home (tn)-Personal clinician-Teams-Coordination/integration-Pt centeredness-QI-Organized/EBM-Access-Leadership-Registries/perf rep-Enhanced asthma care

Intermediate (tn)-Attention to asthma -More data on asthma-Provision of care -More access for in-house tx-Use of spirometry-Patient adherence to tx-Provider guideline concordance-Continuity of care (1o care util)-Smoking status

Practice-level variables (to)-Organizational characteristics -# providers , patients -Practice location-Aggregate patient demographics -race/ethnicity, language -payer mix

Patient-level variables (to)-Patient demographics -Age -Race & Ethnicity -SES (family income, size) -Gender (maybe) -Urban/rural

Patient-level variables (to)-Risk for poor asthma control -Prior exacerbations -Prior utilization -Prior ACT scores-Asthma severity -Persistent/Intermittent (meds)

Patient-level variables (t1)-Risk for poor asthma control -comorbidities (allergy, rhinitis/sinusitis, GERD)

Simplified DAG

Asthma Control-Asthma Control Test (ACT)-Acute asthma exacerbations

Practice Medical Home Characteristics Intermediate Variables

-Patient adherence to treatment plan-Provider guideline concordance-Continuity of care -Smoking cessation

Patient demographics -Age -Race & ethnicity -Socioeconomic status -Urban/rural

Risk for poor asthma control -Prior exacerbations -Prior ACT scores -Comorbidities -Asthma severity

Simplified DAG

Asthma Control-Asthma Control Test (ACT)-Acute asthma exacerbations

Practice Medical Home Characteristics Intermediate Variables

-Patient adherence to treatment plan-Provider guideline concordance-Continuity of care -Smoking cessation

Patient demographics -Age -Race & ethnicity -Socioeconomic status -Urban/rural

Risk for poor asthma control -Prior exacerbations -Prior ACT scores -Comorbidities -Asthma severity

24-month exposure time period

24-month exposure time period

Exposure to practice = 1+ encounter

Covariates: medications, ACT scores, exacerbations, comorbidities

index date

ACT

Repeated measures(7/1/12 – 12/31/13)

Presence/absence of asthma exacerbation during each month of outcome period

Figure 1: Timing for Assessment of Exposure, Outcomes and Covariates

Preliminary analyses• Which domains of medical home show the greatest variability?• Which domains are primarily responsible for the variability in total

medical home score? • What alternative explanations for effects of medical home on

asthma control can be ruled out prior to model building phase?– Information bias (Do medical homes have more ACT data? Greater % of

asthma pts w/ ACT scores? Evidence of heterogeneity of variance in ACT scores?)

– Selection bias (Do medical homes care for certain demographics more than others? Are they more likely to be located in urban vs rural locations? Do they care for more patients, have more providers, have more/less diverse payer mix?)

• Variance components: proportion of variance in outcomes attributed to practice, provider, patient

Model Building1. Bivariate relationships (medical home scores and asthma

control)2. Adjusting for patient demographics

a. Selection process? Elimination process? (only include those with significant bivariate relationships with MH score and asthma control?)

3. Adjusting for practice demographics (provider FTE, payer mix)4. Adjusting for baseline asthma risk (prior exacerbations) and

severity (persistent/intermittent)5. Adjusting for follow-up asthma risk (comorbidities)6. Adjusting for practice processes (potential mediators: use of

spirometry, smoking status, continuity of care, primary care utilization at follow-up)

7. Sensitivity analyses

Sensitivity Analyses

• If claims data available– Asthma exacerbations w/ and w/o hospital/ED– Can answer question, “Compared to non-medical

homes, do medical homes show evidence of increased rates of exacerbations when using only clinical/administrative data, but similar or even lower rates when including claims data?”• Potential to bias effect of medical homes on asthma

control towards worse outcomes, due to differences in access to primary care

Sensitivity Analyses• Assignment to practice

– Linking patient outcomes with appropriate medical home exposure – have we properly identified which practice is responsible for any given ACT value or asthma exacerbation?

– Misclassification bias– Assignment to practice based on utilization 18 months prior AND 6

months inclusive of block• Could bias effect of medical home towards worse outcomes if increased

access for patients who have not been able to get in elsewhere– Assignment to practice based on utilization 18 months prior to block

only • Potential to limit sample size if first visit is exacerbation event; may also bias

non-medical homes towards better outcomes (excluding more exacerbations, lower ACT scores because no prior visits, but lack of visits is reason for poor control)

Tables of asthma results• Separately for child and adult cohorts• Descriptive: patient demographics

– Break down into men/women?• Descriptive: practice demographics• Descriptive: outcomes• Descriptive: medical home total & domain scores• Bivariate analysis (simple relationship between MH and asthma

outcomes)• Bivariate analysis: relationships among covariates• Hierarchical models• Sensitivity analyses