First, I want to direct your attention here: Common trends with Difference in differences. Responses to this question may obliquely address your concerns.
The short answer to your question is, no. Before adopting a difference-in-differences (DD) approach you should start with a visual inspection of pre-treatment trends in your outcome across treatment and control groups. DD designs always set up an implicit treatment and control group comparison using aggregate data. Typically, you want to plot the unconditional evolution of group averages over time.
Here is the canonical DD setup with two groups and two periods:
$$
y_{ist} = \alpha + \gamma T_{s} + \lambda d_{t} + \delta(T_{s} \cdot d_{t}) + \epsilon_{ist},
$$
where we may observe individual/entity $i$, in state $s$, and time period $t$. The generalization of the foregoing equation would include dummies for each state and period but is otherwise unchanged. For example,
$$
y_{ist} = \gamma_{s} + \lambda_{t} + \delta D_{st} + \epsilon_{ist},
$$
where the new treatment dummy $D_{st}$ is the same as before $(T_{s} \cdot d_{t})$. In most settings, the regressor of interest varies only at a more aggregate or group level, such as the state level in this example. Only information at the group level is required for identification of your treatment effect. However, you could incorporate individual level control variables to reduce the residual variance. See the following for a detailed description of using individual level controls in a regression formulation: Control variables- Difference in Difference.
DD is akin to the standard fixed effects estimator. Recall fixed effects assumes the following:
$$
E\left[Y_{0it} \vert i,t \right] = \alpha_{i} + \lambda_{t},
$$
whereas the DD approach assumes:
$$
E\left[Y_{0ist} \vert s,t \right] = \gamma_{s} + \lambda_{t}.
$$
Note, we condition on a group instead of an individual level effect. In essence, this equation states that in the absence of treatment, potential outcomes are determined by the sum of a time-invariant state (group) effect and a time effect that is common across all states. This is sufficient for any treatment/policy/intervention occurring at the state-time level.
In practice, you investigate common trends by gathering data on multiple time periods. In most papers, demonstrating trend equivalence is achieved visually with a good plot of the group outcomes over time before some treatment begins. A visually clear parallelism should be present in the outcomes prior to treatment exposure. If group trends markedly diverge pre-treatment, you could test for group differences with a specification test. I have seen coefficient plots of interactions between the treatment indicator and time dummies to show how effects grow or fade over time.
In sum, plotting the unconditional evolution of outcomes across treatment and control groups is sufficient. A picture is worth a thousand words.
I will let someone else tender their opinion regarding adjustments of these group averages. Hope this helps.
