Coverage Bias in Traditional Telephone Surveys of Low-Income and Young Adults Centers for Disease Control and Prevention National Center for Health Statistics Stephen Blumberg Julian Luke

Percentage of U.S. Households Without Landline Telephones 15.8% of households have only wireless telephones Based on National Health Interview Survey data Based on National Health Interview Survey data

Prevalence of Wireless-Only Adults by Age 30.6% of % of % of % of % of 65+

Prevalence of Wireless-Only Adults by Household Poverty Status 27.4% of adults in poverty 20.8% of adults near poverty 11.9% of higher income adults

Goals of the POQ Article Biases for subgroups of the population may not be adequately reflected in observed biases for the overall population Biases for subgroups of the population may not be adequately reflected in observed biases for the overall population Examined the potential bias that may result when health surveys about young adults and low-income adults exclude households without landline telephones Examined the potential bias that may result when health surveys about young adults and low-income adults exclude households without landline telephones

National Health Interview Survey In-person survey of the civilian noninstitutionalized U.S. population In-person survey of the civilian noninstitutionalized U.S. population Annual household-level response rates are very high (86-92%) Annual household-level response rates are very high (86-92%) Includes questions on residential telephone numbers to permit recontact of participants Includes questions on residential telephone numbers to permit recontact of participants 2003: Added questions on working cellular telephones 2003: Added questions on working cellular telephones 2007: Added questions on relative frequency of calls received on landlines and cell phones 2007: Added questions on relative frequency of calls received on landlines and cell phones

Coverage Bias Two factors determine the degree of coverage bias due to telephone ownership in a telephone survey: Two factors determine the degree of coverage bias due to telephone ownership in a telephone survey: The percentage of persons without landline telephones in the population of interest The percentage of persons without landline telephones in the population of interest The magnitude of the difference between persons with and without landline telephones for the variable of interest The magnitude of the difference between persons with and without landline telephones for the variable of interest

Health Characteristics Examined Health-related behaviors Health-related behaviors 5+ alcoholic drinks in one day (past year) Smoking (current) Leisure-time physical activity (regularly) Health status Health status Excellent or very good health status Serious psychological distress (past 30 days) Obesity Asthma episode (past year) Diabetes (ever diagnosed) Health care service use Health care service use Has a usual place to go for medical care Received influenza vaccine (past year) Tested for HIV (ever) Financial barrier to needed care (past year) Uninsured (current) For these 13 estimates, preliminary weighted data from preliminary weighted data from July – December 2007 were produced by the NHIS Early Release Program.

Percent of U.S. Adults with Various Health Characteristics, by Phone Status Has a landline telephone Wireless- only No telephone 5+ alcoholic drinks in 1 day Current smoker Psychological distress Health excellent / very good Ever diagnosed with diabetes Regular physical activity July – December 2007

Percent of U.S. Adults with Various Health Characteristics, by Phone Status Has a landline telephone Wireless- only No telephone Uninsured (when interviewed) Financial barriers to care Has a usual place for care Flu vaccination Ever tested for HIV July – December 2007

Half (49%) of all wireless-only adults are less than 30 years of age. Half (49%) of all wireless-only adults are less than 30 years of age. Approximately 40% of all wireless-only adults are living in households with income below 200% of the Federal Poverty Level. Approximately 40% of all wireless-only adults are living in households with income below 200% of the Federal Poverty Level. Prevalence of Young Adults and Low-Income Adults Among Wireless-Only Adults

Percent of Young Adults with Various Health Characteristics, by Phone Status Has a landline telephone Wireless- only Health excellent / very good Psychological distress Obese Asthma episode, past year Ever diagnosed with diabetes January – December 2007

Percent of Young Adults with Various Health Characteristics, by Phone Status Has a landline telephone Wireless- only 5+ alcoholic drinks in 1 day Current smoker Regular physical activity Flu vaccination Ever tested for HIV January – December 2007

Percent of Young Adults with Various Health Characteristics, by Phone Status Has a landline telephone Wireless- only Uninsured (when interviewed) Financial barriers to care Has a usual place for care January – December 2007

Potential Bias (in Percentage Points) if an RDD Survey Only Includes Landlines YoungAdults alcoholic drinks in 1 day – 6.4 Current smoker – 2.5 Regular physical activity – 2.2 Ever tested for HIV – 1.9 Has a usual place for care Uninsured (when interviewed) – 2.2 January – December 2007

Potential Bias (in Percentage Points) if an RDD Survey Only Includes Landlines YoungAdults18-29 Statistically significant bias after controlling for demographic characteristics 5+ alcoholic drinks in 1 day – 6.4 – 4.9 Current smoker – 2.5 Regular physical activity – 2.2 – 1.9 Ever tested for HIV – 1.9 – 1.7 Has a usual place for care Uninsured (when interviewed) – 2.2 – 1.7 January – December 2007

Potential Bias (in Percentage Points) if an RDD Survey Only Includes Landlines Low- Income Adults Statistically significant bias after controlling for demographic characteristics 5+ alcoholic drinks in 1 day – 4.9 – 3.0 Current smoker – 2.4 – 1.5 Regular physical activity – 2.4 – 0.9 Ever tested for HIV – 2.8 – 1.4 Has a usual place for care Uninsured (when interviewed) – 3.4 – 1.4 January – December 2007

Conclusions The increase in the prevalence of wireless- only adults has led to nonnegligible coverage biases in landline telephone surveys even after adjusting for demographic differences The increase in the prevalence of wireless- only adults has led to nonnegligible coverage biases in landline telephone surveys even after adjusting for demographic differences Carefully developed sample weights, using multiple demographic control totals, can attenuate the magnitude of bias Carefully developed sample weights, using multiple demographic control totals, can attenuate the magnitude of bias But can other statistical adjustments reduce the bias even further? But can other statistical adjustments reduce the bias even further?

Percent Distribution of Household Telephone Status for Adults, July-December 2007 Wireless Only: 14.5% Landline with Some Wireless: 49.2% Landline Only: 19.1% Unknown: 1.3% Phoneless: 1.9% Wireless Mostly: 14.0%

Percent of U.S. Adults with Various Health Characteristics, by Phone Status Wireless Some Wireless Mostly Wireless Only 5+ alcoholic drinks in 1 day 18.5 < 25.4 < 37.3 Current smoker 16.4 < 20.5 < 30.6 Regular physical activity 31.6 < 36.5 < Ever tested for HIV 34.5 < 45.0 < No usual place for care 10.3 < 17.7 < 32.0 Uninsured11.2 < 16.9 < 28.7 NHIS July - December 2007

Brick, Waksberg, & Keeter (1996) Landline Only Landline and Wireless Wireless Only Phoneless Households with interruptions in service are similar to…

Possible Statistical Adjustment Landline Only Landline and Wireless Wireless Only Phoneless Households with interruptions in service are similar to… Wireless-mostly HHs are similar to…

Future Steps Wireless-mostly adults are similar in many respects to wireless-only adults Wireless-mostly adults are similar in many respects to wireless-only adults But…Are wireless-mostly adults who will respond to a landline survey similar to wireless-only adults? But…Are wireless-mostly adults who will respond to a landline survey similar to wireless-only adults? 14.5% of adults were wireless-only in July- December % of adults were wireless-only in July- December 2007 But…What percent of adults in your surveys subnational population are wireless-only? But…What percent of adults in your surveys subnational population are wireless-only?