What separates a risk assessment an inspector accepts from one they do not

The five steps, and where each one goes wrong

Risk assessment in England still follows the same five steps it has for years: identify the hazards, decide who might be harmed and how, evaluate the risk and decide what to do about it, record what you found and put it into practice, then review it. Services rarely fail because they used a different format. They fail inside the third step and the fifth.

Identify the hazards

Generic hazards produce generic controls. A moving and handling assessment that lists back injury as the hazard has said nothing at all. The hazard is the doorway too narrow for the hoist, the dog that gets underfoot, the loose rail on the third stair. Specific hazards are what make the rest of the document usable, and they are what a reader looks at first to decide whether anybody actually went and looked.

Decide who might be harmed and how

Most assessments name the person receiving care and stop there. A district nurse visiting, a cleaner, a relative who provides part of the support, a neighbour with a key: any of them may be in the same situation. Writing them down takes a minute and it changes the controls, because a control that only works when a trained worker is present is no control at all on a Sunday afternoon.

Evaluate the risk and decide what to do

This is where the two piles separate. A control measure has to be something a named person does, at a stated frequency, with some way of telling whether it happened. Staff to be vigilant is not a control. Staff to be aware of the risk is not a control. Check the rail is secure at the start of every visit and report it the same day if it is not, is a control, because somebody can tell you tomorrow whether it was done.

Record what you found and put it into practice

The recording half is usually fine. The implementation half is where the evidence runs out. If the assessment says a piece of equipment is needed, there should be a date it arrived. If it says two workers are needed, the rota should show two. An assessment whose controls are contradicted by the roster is worse than none, because it proves the service knew.

Review it

Reviews driven only by a calendar date get signed and re-dated with nothing changed. The reviews that count are the ones triggered by something: a fall, a hospital admission, a change of medicines, a new member of staff, a house move, a refusal that was not there before. A review record that says no change after an incident has to explain itself.

What an inspector is testing

• Whether the person, or somebody who speaks for them, had any part in it. A risk assessment written about someone without ever asking them reads as a document about liability.
• Whether it balances risk against the life the person wants. Removing every risk is not the goal, and where the person has capacity to decide otherwise it is not lawful either.
• Whether the controls appear in the care plan, so the worker doing the task can see them without reading the assessment itself.
• Whether the version the worker holds matches the version in the office.
• Whether anything has ever been assessed as too high and stopped. A service that has never restricted anything has never really assessed anything.

Scoring, and what it is for

Numerical scoring is useful for sorting a list and is often mistaken for the assessment itself. A score tells nobody what to do. Keep it if it drives the order in which things get fixed. Drop it if it produces arguments about whether something is a three or a four. What matters after the score is the residual position: what the risk is once the controls are working, and whether the service is content to live with that.

If yours would not survive a look

1. Take three assessments for people you support and read only the control measures. Cross out every one that nobody could verify tomorrow.
2. Rewrite those as an action with a person and a frequency attached.
3. Check the care plan says the same thing. Where it does not, decide which is right and change the other.
4. Look at the review dates. Find one where something happened and the assessment was not touched, and start there.
5. Ask one person you support what they think the risk is, and put their answer in the document in their own words.

The pattern across services that get this right is not a better template. It is that somebody reads the assessments as a reader rather than as an author, and asks what they would do on a Tuesday evening with only the information in front of them.