K10

K10 in Population Surveys and Mental Health Economic Evaluation

The Kessler Psychological Distress Scale asks how often, over a recent period of about a month, a person felt a series of negative emotional states such as being nervous, hopeless or worthless, and adds the answers into one score. Kessler and colleagues built the K10 and its six-question subset, the K6, so that general health surveys could monitor the prevalence of non-specific psychological distress, and the K10 has been part of the Australian National Health Survey since 2001. Health economists meet the scale in survey estimates of high distress, which feed burden of illness work and mental health models, and in trials whose K10 results have to be turned into QALYs. Two scoring conventions are in use, one giving totals of 10 to 50 and the other 0 to 40, so the first check on any K10 figure is which convention produced it.

Development of the K10 and K6 in US national surveys

Kessler and colleagues wanted a short scale that general health surveys could use to monitor the prevalence of non-specific psychological distress and its trends. Pilot questions were given in a US national mail survey of 1,401 people, and a reduced set then went into a US national telephone survey of 1,574 people. The ten-question and six-question scales were built from that reduced set using item response theory models.

Against DSM-IV disorders assessed with the SCID, Kessler and colleagues reported areas under the ROC curve of 0.87 to 0.88 for disorders with a Global Assessment of Functioning score of 0 to 70, and 0.95 to 0.96 for disorders with a score of 0 to 50. The scales therefore separate community cases from non-cases best when the disorder is severe.

The authors argued that brevity and these properties made the K10 and K6 suitable for general health surveys, and noted that the scales were already in use in US and Canadian government surveys and in the WHO World Mental Health Surveys. They also proposed that routine inclusion of either scale in clinical studies would create a crosswalk between community and clinical epidemiology.

The developer's site describes the K6 as a truncated form of the K10 in which four questions are deleted, and offers both scales in interviewer-administered and self-administered versions. The Australian Bureau of Statistics (ABS) describes the K10 as developed by Ron Kessler and Dan Mroczek as a short dimensional measure of non-specific psychological distress in the anxiety-depression spectrum, for use in the US National Health Interview Survey.

Questions, response levels and the two scoring conventions

Every K10 question asks how often the person felt a particular way during the recall period, with five response levels running between none of the time and all of the time. The questions on the developer's self-administered K10 form include feeling tired out for no good reason, feeling so restless that the person could not sit still, and feeling worthless, and the K6 form includes feeling nervous and feeling hopeless.

The recall period depends on the version. The developer's self-administered K6 form asks about the past 30 days, the 2007 Australian National Survey of Mental Health and Wellbeing based the K10 on the 30 days before interview, and the ABS National Health Survey asks about the four weeks before interview. The ABS also notes that the National Health Survey version of the K10 differs slightly from the version in the mental health survey, so results from the two surveys need a comparability check before they are set side by side.

The ABS scores each item from 1 for none of the time to 5 for all of the time, which gives totals of 10 to 50, with higher scores meaning more distress. Kessler and colleagues, in a 2003 paper on screening for serious mental illness, reported the K6 on a range of 0 to 24 and the K10 on a range of 0 to 40, which corresponds to items scored 0 to 4. The two totals differ by a constant.

$$S = \sum_{i=1}^{10} r_i \qquad S^{*} = S-10$$

where $S$ is the K10 total with each response $r_i$ scored 1 to 5, giving a range of 10 to 50, and $S^{*}$ is the total when each item is scored 0 to 4, giving a range of 0 to 40. A mean of 20 on the first convention equals a mean of 10 on the second, so a cut point or a mapping function can be applied only on the scale it was built on.

Australian score bands and screening thresholds

Score bands turn the K10 total into categories that surveys report as population percentages. The ABS groups K10 totals into four levels of psychological distress, and the 2007 National Survey of Mental Health and Wellbeing used the same cut-offs as the National Health Survey.

The ABS information paper on the K10 states that its groupings were developed from an amalgam of work by the Clinical Research Unit for Anxiety and Depression, by Andrews and Slade and by Korten, and that there is no agreed standard for determining cut-off points. The bands are therefore a reporting convention, not a validated diagnostic standard. Andrews and Slade, one of those three inputs, set out to establish reference norms for the K10 using the National Survey of Mental Health and Well-Being, comparing every possible K10 score with symptoms, disability, service use and diagnosis, and concluded that the scale was suitable for assessing population morbidity. For the K6, the National Center for Health Statistics defines serious psychological distress in National Health Interview Survey reports as a score of 13 or more.

Screening accuracy decides how well a band or threshold sorts people with and without a disorder. Furukawa and colleagues tested both scales against CIDI and DSM-IV mood and anxiety disorders in the Australian national survey and reported an area under the ROC curve of 0.90 for the K10 and 0.89 for the K6, both significantly above the 0.80 of the GHQ-12. The K6 was more robust to variation between subsamples, and the authors preferred the K6 for screening for any DSM-IV mood or anxiety disorder because of its brevity and consistency, while noting that the K10 may do better for severe disorders.

The K10 in national surveys and what a survey estimate means

A short scale takes little interview time and can be repeated in every survey round, which suits national statistics. The ABS states that the K10 was included in the National Health Survey in 2001, 2004-05, 2007-08, 2011-12 and 2014-15 because it proved a better predictor of depression and anxiety disorders than the other short general measures used in the 1997 mental health survey. The developer's site notes that the K10 is also in the US National Comorbidity Survey Replication and in the WHO World Mental Health surveys.

A survey estimate of the share of adults with high or very high distress is a prevalence of a score range. The share of the population with a diagnosable disorder is a different quantity, and the gap between the two depends on the screening accuracy of the cut point. A burden of illness study that treats a K10 band as a count of people with a disorder misstates the number of cases and the cost of treating them, in a direction set by whether false positives or false negatives dominate at the threshold.

Turning K10 results into QALYs in cost-utility analysis

A cost-utility analysis needs health state values anchored at full health and death. The K10 is a patient-reported outcome measure without preference weights or a value set, so a K10 total ranks distress without valuing it. Drummond and colleagues, and Brazier and colleagues, describe the two standard routes for such a measure, which are to collect a preference-based measure in the same study or to predict utilities with a mapping function.

Evidence on how utility instruments behave against the K10 comes from a large multi-instrument survey. Mihalopoulos and colleagues used individual data from 917 people with self-reported depression in the International Multi-Instrument Comparison Survey and compared five multi-attribute utility instruments with the K10 and the DASS-21. All five instruments discriminated between the levels of severity measured by the two depression measures, and the AQoL-8D, one of the AQoL instruments, had the highest correlation with them and the best goodness-of-fit transformation properties.

Gamst-Klaussen and colleagues then published mapping algorithms that predict EQ-5D-5L utilities from K10 scores, using 917 respondents with self-reported depression and eight country-specific EQ-5D-5L value sets. They compared six regression models, including ordinary least squares, censored least absolute deviation, beta binomial and fractional logistic regression. The fractional logistic model was generally preferred for both the K10 and the DASS-21, with the beta binomial model best for the Japanese value set, and the authors concluded that the algorithms can adequately predict EQ-5D-5L utilities, so trial data on the K10 can be used to estimate QALYs.

Several choices change the economic result when a mapping is used. The K10 scoring convention must match the one in the estimation data, the value set must suit the decision maker, and the estimation sample, people with self-reported depression, may differ from the trial population. A nonlinear model such as fractional logistic regression gives a different answer when applied to a group's mean K10 score than when applied to each person's score and averaged.

Illustrative worked example of a mapped K10 outcome

All numbers in this example are illustrative. The mapping function is invented to show the arithmetic and is not the published algorithm. A trial of a psychological intervention measures the K10 on the 10 to 50 convention at baseline and at 12 months. Both arms start at a mean of 28. At 12 months the intervention arm scores 20 and the control arm 24, and the intervention adds GBP 450 per person to costs.

1. Illustrative mapping function.

$$U = 1.10-0.015 \times S$$

where $U$ is the predicted utility and $S$ is the K10 total on the 10 to 50 convention.

2. Predicted utilities. At baseline 0.015 × 28 = 0.42, so the utility is 1.10 − 0.42 = 0.68. For the intervention arm at 12 months 0.015 × 20 = 0.30 and 1.10 − 0.30 = 0.80. For the control arm 0.015 × 24 = 0.36 and 1.10 − 0.36 = 0.74.

3. QALYs over one year. Joining the two time points with a straight line, the intervention arm gains (0.68 + 0.80) ÷ 2 = 0.74 QALYs and the control arm (0.68 + 0.74) ÷ 2 = 0.71 QALYs, a difference of 0.74 − 0.71 = 0.03 QALYs.

4. Cost per QALY. The incremental cost-effectiveness ratio is 450 ÷ 0.03 = 15,000, or GBP 15,000 per QALY gained.

5. A scoring convention error. Suppose the baseline mean had been reported on the 0 to 40 convention, as 28 − 10 = 18, and fed into the same function. Then 0.015 × 18 = 0.27 and 1.10 − 0.27 = 0.83, a utility 0.83 − 0.68 = 0.15 higher than the correct value for the same people.

The intervention looks good value at GBP 15,000 per QALY, but the result rests entirely on the mapping. If the same 0.15 error applied at both time points in one arm, it would shift that arm's QALYs by 0.15, five times the 0.03 QALY difference between the arms, and could reverse the ranking of the options. Applying the function to arm means is exact only because this function is linear.

How the K10 differs from distress, the PHQ-9 and other outcome measures

The K10 sits beside several concepts and instruments, and confusing them changes how results are pooled or compared. The main distinctions are between the scale and the construct it measures, and between non-specific and disorder-specific questionnaires.

Distress is the clinical outcome, overall emotional suffering broader than specific anxiety or depression symptoms, and the K10 is one instrument that measures it. The PHQ-9 and GAD-7 are separate questionnaires for depression and for anxiety, each with its own cut points, while the K10 combines anxiety and depression content into one non-specific score. The HADS gives separate anxiety and depression scores and is used mainly in physical illness.

The K10 is a non-preference-based measure, while preference-based measures such as the AQoL and the EQ-5D carry value sets and give utilities directly. Psychiatric epidemiology is the field that uses the K10 to estimate the distribution of distress in populations, and a K10 band is one of its screening tools rather than a diagnosis.

Limitations and common misreadings of K10 scores

Most errors with the K10 come from asking a short screening score to do more than its developers designed it for. Each of the following can bias a cost or a QALY estimate.

• Reading a band as a diagnosis. Furukawa and colleagues measured how well the scales screen for DSM-IV disorders, and an area under the curve of 0.90 still leaves false positives and false negatives at any cut point.
• Mixing scoring conventions. A total on the 0 to 40 convention is ten points below the same responses on the 10 to 50 convention, so bands and mapping functions must match the scoring.
• Comparing surveys without checking the version. The ABS notes that its National Health Survey and mental health survey versions differ slightly, and recall periods of four weeks and 30 days both appear in its documents.
• Applying a mapping outside its sample. The published EQ-5D-5L algorithms were estimated in people with self-reported depression, so their fit in other groups, such as people with physical illness, needs its own check.
• Assuming equal accuracy at every severity. Kessler and colleagues reported higher discrimination for disorders with more severe impairment. A further caution follows for trials in mildly distressed groups, where evidence on the scale's responsiveness and minimal clinically important difference in that group is worth seeking before the K10 is used as the main outcome.

Sources

• Andrews G, Slade T. Interpreting scores on the Kessler Psychological Distress Scale (K10). Australian and New Zealand Journal of Public Health. 2001;25(6):494-497. doi:10.1111/j.1467-842X.2001.tb00310.x
• Australian Bureau of Statistics. National Health Survey: Users' Guide, 2014-15 (cat. no. 4363.0). Kessler Psychological Distress Scale-10 (K10). https://www.abs.gov.au/ausstats/[email protected]/Lookup/by%20Subject/4363.0~2014-15~Main%20Features~Kessler%20Psychological%20Distress%20Scale-10%20(K10)~35
• Australian Bureau of Statistics. Information Paper: Use of the Kessler Psychological Distress Scale in ABS Health Surveys, Australia, 2007-08 (cat. no. 4817.0.55.001). https://www.abs.gov.au/ausstats/[email protected]/lookup/4817.0.55.001Chapter92007-08
• Australian Bureau of Statistics. National Survey of Mental Health and Wellbeing: Summary of Results methodology, 2007. https://www.abs.gov.au/methodologies/national-survey-mental-health-and-wellbeing-summary-results-methodology/2007
• Australian Bureau of Statistics. National Survey of Mental Health and Wellbeing: Users' Guide, 2007 (cat. no. 4327.0). https://www.abs.gov.au/ausstats/[email protected]/Products/4327.0~2007~Main+Features~Interpretation+of+results
• Brazier J, Ratcliffe J, Salomon JA, Tsuchiya A. Measuring and Valuing Health Benefits for Economic Evaluation. 2nd ed. Oxford University Press; 2017.
• Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. 4th ed. Oxford University Press; 2015.
• Furukawa TA, Kessler RC, Slade T, Andrews G. The performance of the K6 and K10 screening scales for psychological distress in the Australian National Survey of Mental Health and Well-Being. Psychological Medicine. 2003;33(2):357-362. https://pubmed.ncbi.nlm.nih.gov/12622315/
• Gamst-Klaussen T, Lamu AN, Chen G, Olsen JA. Assessment of outcome measures for cost-utility analysis in depression: mapping depression scales onto the EQ-5D-5L. BJPsych Open. 2018;4(4):160-166. doi:10.1192/bjo.2018.21
• Kessler RC. K10 and K6 Scales. Harvard Medical School, National Comorbidity Survey. https://www.hcp.med.harvard.edu/ncs/k6_scales.php
• Kessler RC, Andrews G, Colpe LJ, et al. Short screening scales to monitor population prevalences and trends in non-specific psychological distress. Psychological Medicine. 2002;32(6):959-976. doi:10.1017/S0033291702006074
• Kessler RC, Barker PR, Colpe LJ, et al. Screening for serious mental illness in the general population. Archives of General Psychiatry. 2003;60(2):184-189. doi:10.1001/archpsyc.60.2.184
• Mihalopoulos C, Chen G, Iezzi A, Khan MA, Richardson J. Assessing outcomes for cost-utility analysis in depression: comparison of five multi-attribute utility instruments with two depression-specific outcome measures. British Journal of Psychiatry. 2014;205(5):390-397. https://pubmed.ncbi.nlm.nih.gov/25257063
• Weissman J, Pratt LA, Miller EA, Parker JD. Serious psychological distress among adults: United States, 2009-2013. NCHS Data Brief No. 203. National Center for Health Statistics; 2015. https://www.cdc.gov/nchs/data/databriefs/db203.pdf