Reading Your Engagement Survey for Mental Health Signals

October 26th, 20269 min read

Reading Your Engagement Survey for Mental Health Signals

In this article

What the survey measures, and what it doesn't
The items that tend to carry signal
Six patterns worth flagging
A worked illustration
Lines not to cross
Questions to take into the results meeting

In a lot of organizations, the fall engagement survey closed a few weeks ago and the results deck is now making its way around the leadership team. The first slide carries the engagement index. The conversation that follows tends to be about whether it moved two points and how the organization compares with the vendor's benchmark.

Further back in the deck, usually in an appendix nobody reaches, sit the items that say the most about how work is affecting people's mental health: whether workloads are manageable, whether people have any say in how they work, whether they feel able to raise a concern, whether they have the energy the job asks of them. Engagement surveys were not built to measure mental health, and they should never be used to identify individuals who might be struggling. Read at the group level and with some discipline, though, they can show where working conditions are wearing people down. For an employer, that is often more useful than any estimate of how many staff have symptoms, because it points at something the employer can change.

What follows covers which items carry signal, six patterns worth flagging, a worked illustration, the lines not to cross, and a set of questions to take into the results meeting.

What the survey measures, and what it doesn't

Engagement instruments measure attitudes toward work: commitment, pride, discretionary effort, confidence in leadership, intention to stay. Some are long-established commercial instruments, with Gallup's Q12 among the best known. Many organizations use a homegrown questionnaire that has been edited a little every year, which matters later when you try to compare trends.

None of them is a clinical screen or a burnout inventory, and none can tell you whether anyone has depression or an anxiety disorder. It helps to keep the WHO's framing in mind here. In ICD-11, burnout is described as an occupational phenomenon resulting from chronic workplace stress that has not been successfully managed, rather than as a medical condition. The useful question, in other words, is about the workplace.

One more caution before reading anything. Engagement and strain are not opposite ends of a single scale. A team can be proud, committed and exhausted at the same time, and the most committed people are sometimes the ones running closest to empty. A healthy engagement score is not evidence that conditions are fine.

The items that tend to carry signal

Most surveys contain some version of the items below, even if the wording differs. The table shows what a low or falling score may point to, and what the item cannot tell you on its own.

Item family	Typical wording	A low or falling score may point to	What it can't tell you
Workload and pace	"I can get my work done in a normal working week"	Sustained overload, thin staffing, competing priorities	Whether the overload is seasonal or structural
Control	"I have a say in how I do my work"	Low job control, a recognized psychosocial hazard	Which decisions people want a voice in
Manager support	"My manager cares about me as a person"	Strained relationships and fewer early conversations	Whether the manager is under the same strain
Speaking up	"I can raise a concern without fear of negative consequences"	Low psychological safety, problems surfacing late	What happened to shape the answer
Fairness and recognition	"People here are treated fairly"	Effort that feels unrewarded	Whether the issue is pay, respect or process
Energy or wellbeing	"I have the energy I need to do my job well"	Fatigue and strain	How much of the strain comes from outside work
Awareness of support	"I know where to get help if I'm struggling"	Weak communication about the EAP and benefits	Whether people trust the help enough to use it
Intent to stay	"I see myself working here in two years"	Exit driven by disengagement or exhaustion	How much is pull from the job market

If your survey has none of the energy or support items, note it for next year. Adding two or three well-chosen items costs little, provided the wording then stays fixed so the trend means something.

The US Surgeon General's Framework for Workplace Mental Health and Well-Being makes a useful cross-check. Its five essentials (protection from harm, connection and community, work-life harmony, mattering at work, and opportunity for growth) map reasonably well onto standard engagement items, and laying your items against them quickly shows which areas your survey cannot see at all.

Reading list

The Fearless Organization
Amy C. Edmondson
Amazon page
Burnout: The Secret to Unlocking the Stress Cycle
Emily Nagoski and Amelia Nagoski
Amazon page
Lost Connections
Johann Hari
Amazon page

See the full reading list

As an Amazon Associate I earn from qualifying purchases.

Six patterns worth flagging

A healthy average over a heavy lower tail

Percent favorable hides the shape of the distribution. Two teams can both score 70 percent favorable on workload, one with the remaining 30 percent spread across neutral and mild disagreement, the other with a solid block answering "strongly disagree." Ask your vendor for the bottom-box share on each key item. A growing bottom box is often the first sign of trouble, and it can appear a year before the favorable score moves.

Strain falling while commitment holds

When workload and energy items decline but pride and commitment stay level, people are usually absorbing pressure out of loyalty to the team or the work. The engagement index will report that all is well. This is the pattern that tends to come before visible burnout in committed teams, and it is worth pairing with what managers are seeing day to day, using a reference such as the early warning signs of burnout.

Poor workload scores and weak manager support together

Occupational health researchers have studied the demand-control-support model for decades, and its core finding is that high demands combined with low control and low support are a particularly harmful mix. When a unit scores poorly on both workload and manager support, put it near the top of the list. Either problem alone gives you something to work with. Together, they leave people with no buffer.

Silence

A falling response rate in a unit that used to respond well is information. So are answers bunching at neutral, and comment boxes that used to be full going quiet. Common causes include fatigue with surveys that never led to change, doubts about confidentiality, and plain lack of time. Each of those is worth knowing about, and none of them shows up in the engagement index.

Gaps between groups

Compare shifts, sites, job levels, tenure bands, and on-site versus remote staff, always within your minimum group size. First-year employees, night shifts and frontline roles often sit well below the company average on the strain items while the overall figure looks fine. A gap does not explain itself, but it tells you where to go and ask.

Comments in the language of depletion

Free-text comments are where people say what the scaled items cannot. Rather than pulling vivid quotes for the deck, build a simple coding frame of six to eight themes (workload, staffing, scheduling, manager behavior, recognition, tools and systems, wellbeing) and count how often each appears by unit. Have two people code the same sample of 50 comments and compare. If they disagree often, tighten the theme definitions before coding the rest. Counts of comments using phrases such as "can't keep up," "no time to" or "covering for" are a reasonable proxy for strain, and they are far less likely to expose an individual than a quotation is.

A worked illustration

Picture a hypothetical 480-person regional credit union with fourteen branches and a central contact center. The overall engagement index rose by a point this year, and the first slide is upbeat.

Inside the contact center, the engagement score is flat at 68 percent favorable. But agreement with "I can get my work done in a normal working week" has dropped from 61 to 47 percent. The bottom-box share on the energy item has roughly doubled. The response rate fell from 84 to 66 percent, and coded comments mentioning back-to-back calls, or no recovery time between them, went from a handful to more than twenty. The manager-support item held steady.

Read together, this is a committed team under a workload problem, with some signs that people have stopped expecting the survey to change anything. The steady manager score suggests supervisors are seen as being on their staff's side, which makes them the right people to involve in a fix. Sensible next steps would be a look at staffing against call volume, adherence targets and wrap-up time; a short pulse in eight weeks repeating the same three items; and a note to the contact center within a month saying what was heard and what will be tried first.

What the data does not support is a resilience webinar for agents. That would treat a staffing problem as a coping problem, and the people on the phones would notice.

Lines not to cross

Reading for mental health signals raises the stakes on privacy, so set the rules before anyone opens the data.

Group level only. Many engagement surveys are confidential rather than anonymous, meaning the vendor can link responses to HR demographics. Nobody should use that link to look at individuals, and managers should never be invited to guess who wrote a comment.
Minimum group sizes. Five is a common floor and ten is safer for anything sensitive. Watch for stacked filters (site by shift by tenure) that let someone work out a small group's answers by subtraction.
No clinical items. Do not bolt depression or anxiety screening questions onto an engagement survey. It changes what the data is, it can raise questions under the ADA's limits on disability-related inquiries, and it creates information an employer has no good way to act on. The wider considerations are set out in what employers can and can't do on screening. This is general information rather than legal advice, so talk to counsel before changing what you ask.
Conditions, not culprits. Present findings as facts about work design and climate. A team with poor scores is telling you something about its conditions, and it should not come away feeling it has been graded.

Questions to take into the results meeting

Where did the bottom-box share grow, even if favorable scores held steady?
Which units are low on both workload and manager support?
Where did response rates fall, and what happened to last year's feedback in those units?
Are strain items and commitment items moving together or apart?
Which comment themes grew by count, and where?
Which group differences are larger than noise? In a group of 15, a ten-point swing can be two people.
What will change in how work is designed or resourced, not only in what is communicated?
What will employees hear back, from whom, and by what date?

The survey shows how people experience their work. It says much less about what the organization actually has in place, and a quick way to fill that gap is our workplace mental health self-check, which scores four areas in a few minutes. For the climate side specifically, the actions under culture and climate in the four key areas pair well with whatever the strain items turned up, and the resources library lists free survey instruments if you decide to add items next year. Whatever you add, keep the wording identical from one year to the next. A clean trend line is worth more than one perfect question.