About the evidence and original analysis | Who Looks After You?

This book uses a focused selection of primary studies and official sources to test questions about sustained adult care. It is not a systematic review of all caregiving research. The studies differ in country, population, service and outcome. Their differences limit what can be combined.

A trial of nurse calls after hospital discharge does not evaluate every form of support. A trial adding support to an existing day service does not compare respite with no respite. A study of care hours does not measure a person’s whole well-being. Qualitative interviews can reveal a disagreement that a one-sided account misses; they do not establish how common that disagreement is. These limits shape the argument rather than appearing only as qualifications at the end.

The named scenes and dialogue are imagined. The poems and conceptual figures are original. The suggested planning questions are discussion aids, not tested clinical programs or guarantees of safety. The book does not provide medication instructions, capacity assessments or individual benefit determinations. Current service and legal examples were checked in October 2026 and remain jurisdiction-specific.

How the ATUS comparison was made

Chapter 2 analyzes public American Time Use Survey files for 2023 and 2024, supplied by the U.S. Bureau of Labor Statistics. Each respondent describes one day. The underlying survey represents the civilian noninstitutional population aged 15 and older; weekends are oversampled, so raw unweighted averages would be misleading.

The calculation began with 16,217 respondents. To match the eldercare release, a provider had to report unpaid care more than once in the preceding three to four months for an aging-related condition, and have at least one recipient aged 65 or older. This identified 2,554 providers. The age-65 screen selects providers, not recipient-specific care episodes. A qualifying provider may also assist other recipients; the recorded minutes cannot all be assigned to recipients aged 65 and older. Six published BLS benchmarks reproduced at their stated precision before the analysis population was changed.

For the book’s substantive comparison, 20 providers aged 15–17 were removed. Six adults with an unknown diary-care response were excluded instead of treating their unreported time as zero. The final sample was 2,528. Including those six as zero changes the adult mean from 1.09768 to 1.09586 hours, a small difference here. Officially edited demographic values were retained.

Employment includes people temporarily absent from their job. “Not employed” includes both unemployed people and those outside the labor force. An own resident child means the respondent’s own child under 18 in the household, verified against the roster. It does not mean every child in the home; absence of such a child does not mean a person is child-free.

The point estimates use the final person weight. Sampling variance uses all 160 replicate weights, with corresponding replicate numbers aligned across years, following the ATUS guide. Variance is four divided by 160, multiplied by the sum of squared differences between replicate and full-weight estimates. The displayed 95 percent intervals are the estimate plus or minus 1.96 standard errors. They describe sampling uncertainty, not every possible source of error. Population benchmark counts divide the combined person-day weights by the 731 days in 2023–2024; common rescaling cancels in weighted means.

All-provider-day means include valid zero-care days. Care-day means include only diaries with positive recorded eldercare minutes. Thirty-four adults answered yes to care on the diary day but had zero eligible recorded minutes after processing; they remain in the all-day denominator only. The employed group with an own resident child has 356 diaries, of which 53 contain positive care time. The figure preserves its wide uncertainty interval.

The analysis first examined whether employment, parenthood and frequency could be compared together. Some cells were too small, so the main figure uses three broader groups. The original screening rules require at least 100 all-day diaries, 30 positive-activity diaries and a coefficient of variation no greater than 30 percent; conditional estimates need at least 30 positive-care diaries and a coefficient of variation no greater than 30 percent. These are transparent analyst choices, not a claim to reproduce BLS publication rules. The supporting files retain suppression flags and the more detailed cell audit.

Eldercare is an additional tag on an activity. It can overlap leisure, such as companionship. The chart therefore does not stack eldercare with other activities into a false 24-hour total. Leisure and sports include related travel; the broader work measure includes work-related activity and travel, including job search. The sleeping category includes sleeplessness as well as sleep. None is a direct measure of rest, uninterrupted free time, mental load or sleep quality.

The groups differ substantially in age and other characteristics. No causal adjustment is claimed. The differences cannot establish what caregiving, parenthood or employment caused, and the diary cannot reconstruct an individual’s usual week. Exploratory confidence intervals are not adjusted for multiple comparisons.

An independent analyst rebuilt the key calculations with separate code. All 56 selected result rows and six benchmarks reproduced, along with joins, subgroup definitions, uncertainty calculations, the figure and its accessible description. Reproducible scripts, exact input URLs and hashes, and aggregate results accompany the private draft. Independent verification was performed separately. No restricted NSOC/NHATS data were obtained or used.

Public data sources

BLS publications are generally public domain except marked third-party material and protected marks. This book reproduces no protected journal figures, tables, questionnaires or participant quotations. Free access to a paper is not treated as a general republication license. Research summaries are brief and attributed in the chapter notes.